63 Commits

Author SHA1 Message Date
Grail Finder
c2757653a3 Fix: buildable 2026-03-05 14:49:59 +03:00
Grail Finder
4bd6883966 WIP 2026-03-05 14:38:26 +03:00
Grail Finder
7c56e27dbe Dep: trying sugarme tokenizer 2026-03-05 14:27:19 +03:00
Grail Finder
fbc955ca37 Enha: local onnx 2026-03-05 14:13:58 +03:00
Grail Finder
c65c11bcfb Fix: shellmode tab completion 2026-03-05 11:36:35 +03:00
Grail Finder
04f1fd464b Chore: remove cluedo sysprompt 2026-03-05 11:17:01 +03:00
Grail Finder
6e9c453ee0 Enha: explicit app.Draw per textView update for smooth streaming 2026-03-05 10:35:17 +03:00
Grail Finder
645b7351a8 Fix: add different kind of notifiction for fullscreen mode 2026-03-05 09:09:13 +03:00
Grail Finder
57088565bd Fix (notification): being closed by prev notification early 2026-03-05 08:51:04 +03:00
Grail Finder
4b6769e531 Fix (notification): non-blocking way to notify 2026-03-05 08:43:50 +03:00
Grail Finder
d144ee76d9 Chore: pw tools to be disabled as default 2026-03-04 11:45:54 +03:00
Grail Finder
abcaad6609 Enha: native notification implementation 2026-03-04 11:25:13 +03:00
Grail Finder
50ce0200af Fix: graceful shutdown in tui, to avoid other key block 2026-03-04 08:29:47 +03:00
Grail Finder
58ccd63f4a Fix: avoid raw terminal after ctrl+c exit 2026-03-04 08:25:53 +03:00
Grail Finder
3611d7eb59 Fix: missfire of no-vision notification 2026-03-03 16:55:09 +03:00
Grail Finder
8974d2f52c Fix: remove panics from code 2026-03-03 14:51:36 +03:00
Grail Finder
6b0d03f2d6 Fix: decompres before notify 2026-03-03 14:26:06 +03:00
Grail Finder
fb4deb1161 Fix: handle empty choices 2026-03-03 14:13:18 +03:00
Grail Finder
0e5d37666f Enha: id for card map 2026-03-03 11:46:03 +03:00
Grail Finder
093103bdd7 Feat (pw_tools): click_at 2026-03-03 10:53:04 +03:00
Grail Finder
6c9a1ba56b Chore: change 'when askes' to more proactive phrasing 2026-03-03 09:37:34 +03:00
Grail Finder
93ecfc8a34 Enha: palywright dom and elements fetching 2026-03-03 09:27:05 +03:00
Grail Finder
0c9c590d8f Enha (playwright): conditionaly install and use tools 2026-03-03 09:15:18 +03:00
Grail Finder
d130254e88 Chore (pw): restructure 2026-03-03 08:35:18 +03:00
Grail Finder
6e7a063300 Enha: remove window tools if no vision 2026-03-03 08:27:14 +03:00
Grail Finder
c05b93299c Chore: linter complaints 2026-03-03 07:38:57 +03:00
Grail Finder
cad1bd46c1 Feat: playwright tools 2026-03-02 19:20:54 +03:00
Grail Finder
4bddce3700 Enha: compute estimate of non llm text 2026-03-02 15:21:45 +03:00
Grail Finder
fcc71987bf Feat: token use estimation 2026-03-02 14:54:20 +03:00
Grail Finder
8458edf5a8 Enha: interrupt llm and tool both 2026-03-02 12:19:50 +03:00
Grail Finder
07b06bb0d3 Enha: tabcompletion is back in textarea 2026-03-02 12:09:27 +03:00
Grail Finder
3389b1d83b Fix: linter complaints 2026-03-02 11:39:55 +03:00
Grail Finder
4f6000a43a Enha: check if model has vision before giving it vision tools 2026-03-02 11:25:20 +03:00
Grail Finder
9ba46b40cc Feat: screencapture for completion 2026-03-02 11:12:04 +03:00
Grail Finder
5bb456272e Feat: capture window (screenshot) 2026-03-02 10:33:41 +03:00
Grail Finder
8999f48fb9 Fix (completion): handle multiple images in history 2026-03-02 09:23:22 +03:00
Grail Finder
b2f280a7f1 Feat: read img for completion 2026-03-02 07:46:08 +03:00
Grail Finder
65cbd5d6a6 Fix (ctrl+v): trim loaded mark from the model 2026-03-02 07:19:21 +03:00
Grail Finder
caac1d397a Feat: read img tool for chat endpoint 2026-03-02 07:12:28 +03:00
Grail Finder
742f1ca838 Enha: modal affirmation popup on sending empty msg 2026-03-01 16:21:18 +03:00
Grail Finder
e36bade353 Fix: escape with empty textarea not generating response 2026-03-01 13:33:25 +03:00
Grail Finder
01d8bcdbf5 Enha: avoid \n\n in tool collapse 2026-03-01 12:28:23 +03:00
Grail Finder
f6a395bce9 Fix: todo_update 2026-03-01 12:16:17 +03:00
Grail Finder
dc34c63256 Feat: handle llm's cd use 2026-03-01 11:44:43 +03:00
Grail Finder
cdfccf9a24 Enha (llama.cpp): show loaded model on startup 2026-03-01 08:22:02 +03:00
Grail Finder
1f112259d2 Enha(tools.todo): always provide whole todo list 2026-03-01 07:01:13 +03:00
Grail Finder
a505ffaaa9 Fix (tool): handle subcommands 2026-02-28 16:16:32 +03:00
Grail Finder
32be271aa3 Feat (tools): file_edit 2026-02-28 15:40:52 +03:00
Grail Finder
133ec27938 Feat(shell): cd and pipes support 2026-02-28 13:59:54 +03:00
Grail Finder
d79760a289 Fix: do not delete tool calls or lose them on copy 2026-02-28 10:23:03 +03:00
Grail Finder
2580360f91 Fix: removed code that deletes tool calls 2026-02-28 09:13:05 +03:00
Grail Finder
fe4dd0c982 Enha: add go to allowed commands 2026-02-28 08:39:13 +03:00
Grail Finder
83f99d3577 Enha: first chat name convention 2026-02-28 08:09:56 +03:00
Grail Finder
e521434073 Refactor: move msg totext method to main package
logic requires reference to config
2026-02-28 07:57:49 +03:00
Grail Finder
916c5d3904 Enha: icon for collapsed tools 2026-02-27 21:25:26 +03:00
Grail Finder
5b1cbb46fa Chore: linter complaints 2026-02-27 20:03:47 +03:00
Grail Finder
1fcab8365e Enha: tool filter 2026-02-27 18:45:59 +03:00
Grail Finder
c855c30ae2 Enha: save/load message token stats 2026-02-27 11:23:03 +03:00
Grail Finder
915b029d2c Enha: set work/base dir updates filepicker title 2026-02-27 08:37:13 +03:00
Grail Finder
b599e1ab38 Fix: startnewchat fill created_at 2026-02-27 08:14:41 +03:00
Grail Finder
0d94734090 Enha: tool role index for shellmode 2026-02-27 08:07:55 +03:00
Grail Finder
a0ff384b81 Enha: shellmode within inputfield 2026-02-27 07:58:00 +03:00
Grail Finder
09b5e0d08f Enha: shell mode in filepickerdir 2026-02-26 20:10:00 +03:00
32 changed files with 3176 additions and 1068 deletions

View File

@@ -1,4 +1,4 @@
.PHONY: setconfig run lint install-linters setup-whisper build-whisper download-whisper-model docker-up docker-down docker-logs noextra-run installdelve checkdelve
.PHONY: setconfig run lint lintall install-linters setup-whisper build-whisper download-whisper-model docker-up docker-down docker-logs noextra-run installdelve checkdelve
run: setconfig
go build -tags extra -o gf-lt && ./gf-lt
@@ -25,7 +25,10 @@ install-linters: ## Install additional linters (noblanks)
go install github.com/GrailFinder/noblanks-linter/cmd/noblanks@latest
lint: ## Run linters. Use make install-linters first.
golangci-lint run -c .golangci.yml ./...; noblanks ./...
golangci-lint run -c .golangci.yml ./...
lintall: lint
noblanks ./...
# Whisper STT Setup (in batteries directory)
setup-whisper: build-whisper download-whisper-model

View File

@@ -71,8 +71,8 @@ func (ag *AgentClient) buildRequest(sysprompt, msg string) ([]byte, error) {
// Build prompt for completion endpoints
if isCompletion {
var sb strings.Builder
for _, m := range messages {
sb.WriteString(m.ToPrompt())
for i := range messages {
sb.WriteString(messages[i].ToPrompt())
sb.WriteString("\n")
}
prompt := strings.TrimSpace(sb.String())

448
bot.go
View File

@@ -3,6 +3,7 @@ package main
import (
"bufio"
"bytes"
"compress/gzip"
"context"
"encoding/json"
"fmt"
@@ -63,9 +64,13 @@ var (
"google/gemma-3-27b-it:free",
"meta-llama/llama-3.3-70b-instruct:free",
}
LocalModels = []string{}
LocalModels = []string{}
localModelsData *models.LCPModels
orModelsData *models.ORModels
)
var thinkBlockRE = regexp.MustCompile(`(?s)<think>.*?</think>`)
// parseKnownToTag extracts known_to list from content using configured tag.
// Returns cleaned content and list of character names.
func parseKnownToTag(content string) []string {
@@ -136,6 +141,9 @@ func processMessageTag(msg *models.RoleMsg) *models.RoleMsg {
// filterMessagesForCharacter returns messages visible to the specified character.
// If CharSpecificContextEnabled is false, returns all messages.
func filterMessagesForCharacter(messages []models.RoleMsg, character string) []models.RoleMsg {
if strings.Contains(cfg.CurrentAPI, "chat") {
return messages
}
if cfg == nil || !cfg.CharSpecificContextEnabled || character == "" {
return messages
}
@@ -143,97 +151,67 @@ func filterMessagesForCharacter(messages []models.RoleMsg, character string) []m
return messages
}
filtered := make([]models.RoleMsg, 0, len(messages))
for _, msg := range messages {
for i := range messages {
// If KnownTo is nil or empty, message is visible to all
// system msg cannot be filtered
if len(msg.KnownTo) == 0 || msg.Role == "system" {
filtered = append(filtered, msg)
if len(messages[i].KnownTo) == 0 || messages[i].Role == "system" {
filtered = append(filtered, messages[i])
continue
}
if slices.Contains(msg.KnownTo, character) {
if slices.Contains(messages[i].KnownTo, character) {
// Check if character is in KnownTo lis
filtered = append(filtered, msg)
filtered = append(filtered, messages[i])
}
}
return filtered
}
func cleanToolCalls(messages []models.RoleMsg) []models.RoleMsg {
// If AutoCleanToolCallsFromCtx is false, keep tool call messages in context
if cfg != nil && !cfg.AutoCleanToolCallsFromCtx {
return consolidateAssistantMessages(messages)
}
cleaned := make([]models.RoleMsg, 0, len(messages))
for i, msg := range messages {
// recognize the message as the tool call and remove it
// tool call in last msg should stay
if msg.ToolCallID == "" || i == len(messages)-1 {
cleaned = append(cleaned, msg)
}
}
return consolidateAssistantMessages(cleaned)
}
// consolidateAssistantMessages merges consecutive assistant messages into a single message
func consolidateAssistantMessages(messages []models.RoleMsg) []models.RoleMsg {
if len(messages) == 0 {
return messages
}
consolidated := make([]models.RoleMsg, 0, len(messages))
currentAssistantMsg := models.RoleMsg{}
isBuildingAssistantMsg := false
for i := 0; i < len(messages); i++ {
msg := messages[i]
// assistant role only
if msg.Role == cfg.AssistantRole {
// If this is an assistant message, start or continue building
if !isBuildingAssistantMsg {
// Start accumulating assistant message
currentAssistantMsg = msg.Copy()
isBuildingAssistantMsg = true
} else {
// Continue accumulating - append content to the current assistant message
if currentAssistantMsg.IsContentParts() || msg.IsContentParts() {
// Handle structured content
if !currentAssistantMsg.IsContentParts() {
// Preserve the original ToolCallID before conversion
originalToolCallID := currentAssistantMsg.ToolCallID
// Convert existing content to content parts
currentAssistantMsg = models.NewMultimodalMsg(currentAssistantMsg.Role, []interface{}{models.TextContentPart{Type: "text", Text: currentAssistantMsg.Content}})
// Restore the original ToolCallID to preserve tool call linking
currentAssistantMsg.ToolCallID = originalToolCallID
}
if msg.IsContentParts() {
currentAssistantMsg.ContentParts = append(currentAssistantMsg.ContentParts, msg.GetContentParts()...)
} else if msg.Content != "" {
currentAssistantMsg.AddTextPart(msg.Content)
}
} else {
// Simple string content
if currentAssistantMsg.Content != "" {
currentAssistantMsg.Content += "\n" + msg.Content
} else {
currentAssistantMsg.Content = msg.Content
}
// ToolCallID is already preserved since we're not creating a new message object when just concatenating content
}
result := make([]models.RoleMsg, 0, len(messages))
for i := range messages {
// Non-assistant messages are appended as-is
if messages[i].Role != cfg.AssistantRole {
result = append(result, messages[i])
continue
}
// Assistant message: start a new block or merge with the last one
if len(result) == 0 || result[len(result)-1].Role != cfg.AssistantRole {
// First assistant in a block: append a copy (avoid mutating input)
result = append(result, messages[i].Copy())
continue
}
// Merge with the last assistant message
last := &result[len(result)-1]
// If either message has structured content, unify to ContentParts
if last.IsContentParts() || messages[i].IsContentParts() {
// Convert last to ContentParts if needed, preserving ToolCallID
if !last.IsContentParts() {
toolCallID := last.ToolCallID
*last = models.NewMultimodalMsg(last.Role, []interface{}{
models.TextContentPart{Type: "text", Text: last.Content},
})
last.ToolCallID = toolCallID
}
// Add current message's content to last
if messages[i].IsContentParts() {
last.ContentParts = append(last.ContentParts, messages[i].GetContentParts()...)
} else if messages[i].Content != "" {
last.AddTextPart(messages[i].Content)
}
} else {
// This is not an assistant message
// If we were building an assistant message, add it to the result
if isBuildingAssistantMsg {
consolidated = append(consolidated, currentAssistantMsg)
isBuildingAssistantMsg = false
// Both simple strings: concatenate with newline
if last.Content != "" && messages[i].Content != "" {
last.Content += "\n" + messages[i].Content
} else if messages[i].Content != "" {
last.Content = messages[i].Content
}
// Add the non-assistant message
consolidated = append(consolidated, msg)
// ToolCallID is already preserved in last
}
}
// Don't forget the last assistant message if we were building one
if isBuildingAssistantMsg {
consolidated = append(consolidated, currentAssistantMsg)
}
return consolidated
return result
}
// GetLogLevel returns the current log level as a string
@@ -290,9 +268,7 @@ func warmUpModel() {
// Continue with warmup attempt anyway
}
if loaded {
if err := notifyUser("model already loaded", "Model "+chatBody.Model+" is already loaded."); err != nil {
logger.Debug("failed to notify user", "error", err)
}
showToast("model already loaded", "Model "+chatBody.Model+" is already loaded.")
return
}
go func() {
@@ -380,6 +356,7 @@ func fetchORModels(free bool) ([]string, error) {
if err := json.NewDecoder(resp.Body).Decode(data); err != nil {
return nil, err
}
orModelsData = data
freeModels := data.ListModels(free)
return freeModels, nil
}
@@ -404,22 +381,22 @@ func fetchLCPModels() ([]string, error) {
// fetchLCPModelsWithLoadStatus returns models with "(loaded)" indicator for loaded models
func fetchLCPModelsWithLoadStatus() ([]string, error) {
models, err := fetchLCPModelsWithStatus()
modelList, err := fetchLCPModelsWithStatus()
if err != nil {
return nil, err
}
result := make([]string, 0, len(models.Data))
result := make([]string, 0, len(modelList.Data))
li := 0 // loaded index
for i, m := range models.Data {
for i, m := range modelList.Data {
modelName := m.ID
if m.Status.Value == "loaded" {
modelName = "(loaded) " + modelName
modelName = models.LoadedMark + modelName
li = i
}
result = append(result, modelName)
}
if li == 0 {
return result, nil // no loaded models
return result, nil // no loaded modelList
}
loadedModel := result[li]
result = append(result[:li], result[li+1:]...)
@@ -441,6 +418,7 @@ func fetchLCPModelsWithStatus() (*models.LCPModels, error) {
if err := json.NewDecoder(resp.Body).Decode(data); err != nil {
return nil, err
}
localModelsData = data
return data, nil
}
@@ -458,6 +436,33 @@ func isModelLoaded(modelID string) (bool, error) {
return false, nil
}
func ModelHasVision(api, modelID string) bool {
switch {
case strings.Contains(api, "deepseek"):
return false
case strings.Contains(api, "openrouter"):
resp, err := http.Get("https://openrouter.ai/api/v1/models")
if err != nil {
logger.Warn("failed to fetch OR models for vision check", "error", err)
return false
}
defer resp.Body.Close()
orm := &models.ORModels{}
if err := json.NewDecoder(resp.Body).Decode(orm); err != nil {
logger.Warn("failed to decode OR models for vision check", "error", err)
return false
}
return orm.HasVision(modelID)
default:
models, err := fetchLCPModelsWithStatus()
if err != nil {
logger.Warn("failed to fetch LCP models for vision check", "error", err)
return false
}
return models.HasVision(modelID)
}
}
// monitorModelLoad starts a goroutine that periodically checks if the specified model is loaded.
func monitorModelLoad(modelID string) {
go func() {
@@ -476,9 +481,7 @@ func monitorModelLoad(modelID string) {
continue
}
if loaded {
if err := notifyUser("model loaded", "Model "+modelID+" is now loaded and ready."); err != nil {
logger.Debug("failed to notify user", "error", err)
}
showToast("model loaded", "Model "+modelID+" is now loaded and ready.")
refreshChatDisplay()
return
}
@@ -489,6 +492,17 @@ func monitorModelLoad(modelID string) {
// extractDetailedErrorFromBytes extracts detailed error information from response body bytes
func extractDetailedErrorFromBytes(body []byte, statusCode int) string {
// Try to decompress gzip if the response is compressed
if len(body) >= 2 && body[0] == 0x1f && body[1] == 0x8b {
reader, err := gzip.NewReader(bytes.NewReader(body))
if err == nil {
decompressed, err := io.ReadAll(reader)
reader.Close()
if err == nil {
body = decompressed
}
}
}
// Try to parse as JSON to extract detailed error information
var errorResponse map[string]any
if err := json.Unmarshal(body, &errorResponse); err == nil {
@@ -554,9 +568,7 @@ func sendMsgToLLM(body io.Reader) {
req, err := http.NewRequest("POST", cfg.CurrentAPI, body)
if err != nil {
logger.Error("newreq error", "error", err)
if err := notifyUser("error", "apicall failed:"+err.Error()); err != nil {
logger.Error("failed to notify", "error", err)
}
showToast("error", "apicall failed:"+err.Error())
streamDone <- true
return
}
@@ -568,9 +580,7 @@ func sendMsgToLLM(body io.Reader) {
resp, err := httpClient.Do(req)
if err != nil {
logger.Error("llamacpp api", "error", err)
if err := notifyUser("error", "apicall failed:"+err.Error()); err != nil {
logger.Error("failed to notify", "error", err)
}
showToast("error", "apicall failed:"+err.Error())
streamDone <- true
return
}
@@ -581,9 +591,7 @@ func sendMsgToLLM(body io.Reader) {
if err != nil {
logger.Error("failed to read error response body", "error", err, "status_code", resp.StatusCode)
detailedError := fmt.Sprintf("HTTP Status: %d, Failed to read response body: %v", resp.StatusCode, err)
if err := notifyUser("API Error", detailedError); err != nil {
logger.Error("failed to notify", "error", err)
}
showToast("API Error", detailedError)
resp.Body.Close()
streamDone <- true
return
@@ -591,9 +599,7 @@ func sendMsgToLLM(body io.Reader) {
// Parse the error response for detailed information
detailedError := extractDetailedErrorFromBytes(bodyBytes, resp.StatusCode)
logger.Error("API returned error status", "status_code", resp.StatusCode, "detailed_error", detailedError)
if err := notifyUser("API Error", detailedError); err != nil {
logger.Error("failed to notify", "error", err)
}
showToast("API Error", detailedError)
resp.Body.Close()
streamDone <- true
return
@@ -630,16 +636,12 @@ func sendMsgToLLM(body io.Reader) {
detailedError := fmt.Sprintf("Streaming connection closed unexpectedly (Status: %d). This may indicate an API error. Check your API provider and model settings.", resp.StatusCode)
logger.Error("error reading response body", "error", err, "detailed_error", detailedError,
"status_code", resp.StatusCode, "user_role", cfg.UserRole, "parser", chunkParser, "link", cfg.CurrentAPI)
if err := notifyUser("API Error", detailedError); err != nil {
logger.Error("failed to notify", "error", err)
}
showToast("API Error", detailedError)
} else {
logger.Error("error reading response body", "error", err, "line", string(line),
"user_role", cfg.UserRole, "parser", chunkParser, "link", cfg.CurrentAPI)
// if err.Error() != "EOF" {
if err := notifyUser("API error", err.Error()); err != nil {
logger.Error("failed to notify", "error", err)
}
showToast("API error", err.Error())
}
streamDone <- true
break
@@ -666,9 +668,7 @@ func sendMsgToLLM(body io.Reader) {
if err != nil {
logger.Error("error parsing response body", "error", err,
"line", string(line), "url", cfg.CurrentAPI)
if err := notifyUser("LLM Response Error", "Failed to parse LLM response: "+err.Error()); err != nil {
logger.Error("failed to notify user", "error", err)
}
showToast("LLM Response Error", "Failed to parse LLM response: "+err.Error())
streamDone <- true
break
}
@@ -743,7 +743,7 @@ func sendMsgToLLM(body io.Reader) {
}
interrupt:
if interruptResp { // read bytes, so it would not get into beginning of the next req
interruptResp = false
// interruptResp = false
logger.Info("interrupted bot response", "chunk_counter", counter)
streamDone <- true
break
@@ -777,14 +777,14 @@ func showSpinner() {
botPersona = cfg.WriteNextMsgAsCompletionAgent
}
for botRespMode || toolRunningMode {
time.Sleep(100 * time.Millisecond)
time.Sleep(400 * time.Millisecond)
spin := i % len(spinners)
app.QueueUpdateDraw(func() {
switch {
case toolRunningMode:
textArea.SetTitle(spinners[spin] + " tool")
case botRespMode:
textArea.SetTitle(spinners[spin] + " " + botPersona)
textArea.SetTitle(spinners[spin] + " " + botPersona + " (F6 to interrupt)")
default:
textArea.SetTitle(spinners[spin] + " input")
}
@@ -797,6 +797,7 @@ func showSpinner() {
}
func chatRound(r *models.ChatRoundReq) error {
interruptResp = false
botRespMode = true
go showSpinner()
updateStatusLine()
@@ -960,7 +961,12 @@ out:
if err := updateStorageChat(activeChatName, chatBody.Messages); err != nil {
logger.Warn("failed to update storage", "error", err, "name", activeChatName)
}
if findCall(respText.String(), toolResp.String()) {
// Strip think blocks before parsing for tool calls
respTextNoThink := thinkBlockRE.ReplaceAllString(respText.String(), "")
if interruptResp {
return nil
}
if findCall(respTextNoThink, toolResp.String()) {
return nil
}
// Check if this message was sent privately to specific characters
@@ -982,7 +988,7 @@ func cleanChatBody() {
}
// Tool request cleaning is now configurable via AutoCleanToolCallsFromCtx (default false)
// /completion msg where part meant for user and other part tool call
chatBody.Messages = cleanToolCalls(chatBody.Messages)
// chatBody.Messages = cleanToolCalls(chatBody.Messages)
chatBody.Messages = consolidateAssistantMessages(chatBody.Messages)
}
@@ -1096,22 +1102,38 @@ func findCall(msg, toolCall string) bool {
}
lastToolCall.Args = openAIToolMap
fc = lastToolCall
// Set lastToolCall.ID from parsed tool call ID if available
if len(openAIToolMap) > 0 {
if id, exists := openAIToolMap["id"]; exists {
lastToolCall.ID = id
}
}
// NOTE: We do NOT override lastToolCall.ID from arguments.
// The ID should come from the streaming response (chunk.ToolID) set earlier.
// Some tools like todo_create have "id" in their arguments which is NOT the tool call ID.
} else {
jsStr := toolCallRE.FindString(msg)
if jsStr == "" { // no tool call case
return false
}
prefix := "__tool_call__\n"
suffix := "\n__tool_call__"
jsStr = strings.TrimSuffix(strings.TrimPrefix(jsStr, prefix), suffix)
// Remove prefix/suffix with flexible whitespace handling
jsStr = strings.TrimSpace(jsStr)
jsStr = strings.TrimPrefix(jsStr, "__tool_call__")
jsStr = strings.TrimSuffix(jsStr, "__tool_call__")
jsStr = strings.TrimSpace(jsStr)
// HTML-decode the JSON string to handle encoded characters like &lt; -> <=
decodedJsStr := html.UnescapeString(jsStr)
// Try to find valid JSON bounds (first { to last })
start := strings.Index(decodedJsStr, "{")
end := strings.LastIndex(decodedJsStr, "}")
if start == -1 || end == -1 || end <= start {
logger.Error("failed to find valid JSON in tool call", "json_string", decodedJsStr)
toolResponseMsg := models.RoleMsg{
Role: cfg.ToolRole,
Content: "Error processing tool call: no valid JSON found. Please check the JSON format.",
}
chatBody.Messages = append(chatBody.Messages, toolResponseMsg)
crr := &models.ChatRoundReq{
Role: cfg.AssistantRole,
}
chatRoundChan <- crr
return true
}
decodedJsStr = decodedJsStr[start : end+1]
var err error
fc, err = unmarshalFuncCall(decodedJsStr)
if err != nil {
@@ -1138,14 +1160,18 @@ func findCall(msg, toolCall string) bool {
lastToolCall.Args = fc.Args
}
// we got here => last msg recognized as a tool call (correct or not)
// make sure it has ToolCallID
if chatBody.Messages[len(chatBody.Messages)-1].ToolCallID == "" {
// Tool call IDs should be alphanumeric strings with length 9!
chatBody.Messages[len(chatBody.Messages)-1].ToolCallID = randString(9)
// Use the tool call ID from streaming response (lastToolCall.ID)
// Don't generate random ID - the ID should match between assistant message and tool response
lastMsgIdx := len(chatBody.Messages) - 1
if lastToolCall.ID != "" {
chatBody.Messages[lastMsgIdx].ToolCallID = lastToolCall.ID
}
// Ensure lastToolCall.ID is set, fallback to assistant message's ToolCallID
if lastToolCall.ID == "" {
lastToolCall.ID = chatBody.Messages[len(chatBody.Messages)-1].ToolCallID
// Store tool call info in the assistant message
// Convert Args map to JSON string for storage
chatBody.Messages[lastMsgIdx].ToolCall = &models.ToolCall{
ID: lastToolCall.ID,
Name: lastToolCall.Name,
Args: mapToString(lastToolCall.Args),
}
// call a func
_, ok := fnMap[fc.Name]
@@ -1175,16 +1201,61 @@ func findCall(msg, toolCall string) bool {
toolRunningMode = true
resp := callToolWithAgent(fc.Name, fc.Args)
toolRunningMode = false
toolMsg := string(resp) // Remove the "tool response: " prefix and %+v formatting
toolMsg := string(resp)
logger.Info("llm used a tool call", "tool_name", fc.Name, "too_args", fc.Args, "id", fc.ID, "tool_resp", toolMsg)
fmt.Fprintf(textView, "%s[-:-:b](%d) <%s>: [-:-:-]\n%s\n",
"\n\n", len(chatBody.Messages), cfg.ToolRole, toolMsg)
// Create tool response message with the proper tool_call_id
toolResponseMsg := models.RoleMsg{
Role: cfg.ToolRole,
Content: toolMsg,
ToolCallID: lastToolCall.ID, // Use the stored tool call ID
// Mark shell commands as always visible
isShellCommand := fc.Name == "execute_command"
// Check if response is multimodal content (image)
var toolResponseMsg models.RoleMsg
if strings.HasPrefix(strings.TrimSpace(toolMsg), `{"type":"multimodal_content"`) {
// Parse multimodal content response
multimodalResp := models.MultimodalToolResp{}
if err := json.Unmarshal([]byte(toolMsg), &multimodalResp); err == nil && multimodalResp.Type == "multimodal_content" {
// Create RoleMsg with ContentParts
var contentParts []any
for _, part := range multimodalResp.Parts {
partType := part["type"]
switch partType {
case "text":
contentParts = append(contentParts, models.TextContentPart{Type: "text", Text: part["text"]})
case "image_url":
contentParts = append(contentParts, models.ImageContentPart{
Type: "image_url",
ImageURL: struct {
URL string `json:"url"`
}{URL: part["url"]},
})
default:
continue
}
}
toolResponseMsg = models.RoleMsg{
Role: cfg.ToolRole,
ContentParts: contentParts,
HasContentParts: true,
ToolCallID: lastToolCall.ID,
IsShellCommand: isShellCommand,
}
} else {
// Fallback to regular content
toolResponseMsg = models.RoleMsg{
Role: cfg.ToolRole,
Content: toolMsg,
ToolCallID: lastToolCall.ID,
IsShellCommand: isShellCommand,
}
}
} else {
toolResponseMsg = models.RoleMsg{
Role: cfg.ToolRole,
Content: toolMsg,
ToolCallID: lastToolCall.ID,
IsShellCommand: isShellCommand,
}
}
fmt.Fprintf(textView, "%s[-:-:b](%d) <%s>: [-:-:-]\n%s\n",
"\n\n", len(chatBody.Messages), cfg.ToolRole, toolResponseMsg.GetText())
chatBody.Messages = append(chatBody.Messages, toolResponseMsg)
logger.Debug("findCall: added actual tool response", "role", toolResponseMsg.Role, "content_len", len(toolResponseMsg.Content), "tool_call_id", toolResponseMsg.ToolCallID, "message_count_after_add", len(chatBody.Messages))
// Clear the stored tool call ID after using it
@@ -1200,12 +1271,42 @@ func findCall(msg, toolCall string) bool {
func chatToTextSlice(messages []models.RoleMsg, showSys bool) []string {
resp := make([]string, len(messages))
for i, msg := range messages {
// INFO: skips system msg and tool msg
if !showSys && (msg.Role == cfg.ToolRole || msg.Role == "system") {
for i := range messages {
icon := fmt.Sprintf("[-:-:b](%d) <%s>:[-:-:-]", i, messages[i].Role)
// Handle tool call indicators (assistant messages with tool call but empty content)
if messages[i].Role == cfg.AssistantRole && messages[i].ToolCall != nil && messages[i].ToolCall.ID != "" {
// This is a tool call indicator - show collapsed
if toolCollapsed {
toolName := messages[i].ToolCall.Name
resp[i] = strings.ReplaceAll(fmt.Sprintf("%s\n%s\n[yellow::i][tool call: %s (press Ctrl+T to expand)][-:-:-]\n", icon, messages[i].GetText(), toolName), "\n\n", "\n")
} else {
// Show full tool call info
toolName := messages[i].ToolCall.Name
resp[i] = strings.ReplaceAll(fmt.Sprintf("%s\n%s\n[yellow::i][tool call: %s][-:-:-]\nargs: %s\nid: %s\n", icon, messages[i].GetText(), toolName, messages[i].ToolCall.Args, messages[i].ToolCall.ID), "\n\n", "\n")
}
continue
}
resp[i] = msg.ToText(i)
// Handle tool responses
if messages[i].Role == cfg.ToolRole || messages[i].Role == "tool" {
// Always show shell commands
if messages[i].IsShellCommand {
resp[i] = MsgToText(i, &messages[i])
continue
}
// Hide non-shell tool responses when collapsed
if toolCollapsed {
resp[i] = icon + "\n[yellow::i][tool resp (press Ctrl+T to expand)][-:-:-]\n"
continue
}
// When expanded, show tool responses
resp[i] = MsgToText(i, &messages[i])
continue
}
// INFO: skips system msg when showSys is false
if !showSys && messages[i].Role == "system" {
continue
}
resp[i] = MsgToText(i, &messages[i])
}
return resp
}
@@ -1239,20 +1340,6 @@ func chatToText(messages []models.RoleMsg, showSys bool) string {
return text
}
func removeThinking(chatBody *models.ChatBody) {
msgs := []models.RoleMsg{}
for _, msg := range chatBody.Messages {
// Filter out tool messages and thinking markers
if msg.Role == cfg.ToolRole {
continue
}
// find thinking and remove it - use SetText to preserve ContentParts
msg.SetText(thinkRE.ReplaceAllString(msg.GetText(), ""))
msgs = append(msgs, msg)
}
chatBody.Messages = msgs
}
func addNewChat(chatName string) {
id, err := store.ChatGetMaxID()
if err != nil {
@@ -1289,8 +1376,8 @@ func applyCharCard(cc *models.CharCard, loadHistory bool) {
}
func charToStart(agentName string, keepSysP bool) bool {
cc, ok := sysMap[agentName]
if !ok {
cc := GetCardByRole(agentName)
if cc == nil {
return false
}
applyCharCard(cc, keepSysP)
@@ -1307,11 +1394,28 @@ func updateModelLists() {
}
// if llama.cpp started after gf-lt?
localModelsMu.Lock()
LocalModels, err = fetchLCPModels()
LocalModels, err = fetchLCPModelsWithLoadStatus()
localModelsMu.Unlock()
if err != nil {
logger.Warn("failed to fetch llama.cpp models", "error", err)
}
// set already loaded model in llama.cpp
if strings.Contains(cfg.CurrentAPI, "localhost") || strings.Contains(cfg.CurrentAPI, "127.0.0.1") {
localModelsMu.Lock()
defer localModelsMu.Unlock()
for i := range LocalModels {
if strings.Contains(LocalModels[i], models.LoadedMark) {
m := strings.TrimPrefix(LocalModels[i], models.LoadedMark)
cfg.CurrentModel = m
chatBody.Model = m
cachedModelColor = "green"
updateStatusLine()
updateToolCapabilities()
app.Draw()
return
}
}
}
}
func refreshLocalModelsIfEmpty() {
@@ -1334,15 +1438,15 @@ func refreshLocalModelsIfEmpty() {
func summarizeAndStartNewChat() {
if len(chatBody.Messages) == 0 {
_ = notifyUser("info", "No chat history to summarize")
showToast("info", "No chat history to summarize")
return
}
_ = notifyUser("info", "Summarizing chat history...")
showToast("info", "Summarizing chat history...")
// Call the summarize_chat tool via agent
summaryBytes := callToolWithAgent("summarize_chat", map[string]string{})
summary := string(summaryBytes)
if summary == "" {
_ = notifyUser("error", "Failed to generate summary")
showToast("error", "Failed to generate summary")
return
}
// Start a new chat
@@ -1361,7 +1465,7 @@ func summarizeAndStartNewChat() {
if err := updateStorageChat(activeChatName, chatBody.Messages); err != nil {
logger.Warn("failed to update storage after injecting summary", "error", err)
}
_ = notifyUser("info", "Chat summarized and new chat started with summary as tool response")
showToast("info", "Chat summarized and new chat started with summary as tool response")
}
func init() {
@@ -1374,15 +1478,6 @@ func init() {
os.Exit(1)
return
}
// Set image base directory for path display
baseDir := cfg.FilePickerDir
if baseDir == "" || baseDir == "." {
// Resolve "." to current working directory
if wd, err := os.Getwd(); err == nil {
baseDir = wd
}
}
models.SetImageBaseDir(baseDir)
defaultStarter = []models.RoleMsg{
{Role: "system", Content: basicSysMsg},
{Role: cfg.AssistantRole, Content: defaultFirstMsg},
@@ -1397,8 +1492,6 @@ func init() {
}
// load cards
basicCard.Role = cfg.AssistantRole
// toolCard.Role = cfg.AssistantRole
//
logLevel.Set(slog.LevelInfo)
logger = slog.New(slog.NewTextHandler(logfile, &slog.HandlerOptions{Level: logLevel}))
store = storage.NewProviderSQL(cfg.DBPATH, logger)
@@ -1431,6 +1524,23 @@ func init() {
if cfg.STT_ENABLED {
asr = NewSTT(logger, cfg)
}
if cfg.PlaywrightEnabled {
if err := checkPlaywright(); err != nil {
// slow, need a faster check if playwright install
if err := installPW(); err != nil {
logger.Error("failed to install playwright", "error", err)
cancel()
os.Exit(1)
return
}
if err := checkPlaywright(); err != nil {
logger.Error("failed to run playwright", "error", err)
cancel()
os.Exit(1)
return
}
}
}
// Initialize scrollToEndEnabled based on config
scrollToEndEnabled = cfg.AutoScrollEnabled
go updateModelLists()

View File

@@ -27,7 +27,6 @@ AutoCleanToolCallsFromCtx = false
RAGEnabled = false
RAGBatchSize = 1
RAGWordLimit = 80
RAGWorkers = 2
RAGDir = "ragimport"
# extra tts
TTS_ENABLED = false
@@ -57,3 +56,6 @@ StripThinkingFromAPI = true # Strip <think> blocks from messages before sending
# Valid values: xhigh, high, medium, low, minimal, none (empty or none = disabled)
# Models that support reasoning will include thinking content wrapped in <think> tags
ReasoningEffort = "medium"
# playwright tools
PlaywrightEnabled = false
PlaywrightDebug = false

View File

@@ -39,7 +39,6 @@ type Config struct {
// rag settings
RAGEnabled bool `toml:"RAGEnabled"`
RAGDir string `toml:"RAGDir"`
RAGWorkers uint32 `toml:"RAGWorkers"`
RAGBatchSize int `toml:"RAGBatchSize"`
RAGWordLimit uint32 `toml:"RAGWordLimit"`
// deepseek
@@ -71,6 +70,9 @@ type Config struct {
CharSpecificContextEnabled bool `toml:"CharSpecificContextEnabled"`
CharSpecificContextTag string `toml:"CharSpecificContextTag"`
AutoTurn bool `toml:"AutoTurn"`
// playwright browser
PlaywrightEnabled bool `toml:"PlaywrightEnabled"`
PlaywrightDebug bool `toml:"PlaywrightDebug"` // !headless
}
func LoadConfig(fn string) (*Config, error) {

View File

@@ -80,9 +80,6 @@ This document explains how to set up and configure the application using the `co
#### RAGWordLimit (`80`)
- Maximum number of words in a batch to tokenize and store.
#### RAGWorkers (`2`)
- Number of concurrent workers for RAG processing.
#### RAGDir (`"ragimport"`)
- Directory containing documents for RAG processing.
@@ -165,6 +162,15 @@ Those could be switched in program, but also bould be setup in config.
#### ToolUse
- Enable or disable explanation of tools to llm, so it could use them.
#### Playwright Browser Automation
These settings enable browser automation tools available to the LLM.
- **PlaywrightEnabled** (`false`)
- Enable or disable Playwright browser automation tools for the LLM. When enabled, the LLM can use tools like `pw_browser`, `pw_close`, and `pw_status` to automate browser interactions.
- **PlaywrightDebug** (`false`)
- Enable debug mode for Playwright browser. When set to `true`, the browser runs in visible (non-headless) mode, displaying the GUI for debugging purposes. When `false`, the browser runs in headless mode by default.
### StripThinkingFromAPI (`true`)
- Strip thinking blocks from messages before sending to LLM. Keeps them in chat history for local viewing but reduces token usage in API calls.

11
go.mod
View File

@@ -14,25 +14,36 @@ require (
github.com/jmoiron/sqlx v1.4.0
github.com/ledongthuc/pdf v0.0.0-20250511090121-5959a4027728
github.com/neurosnap/sentences v1.1.2
github.com/playwright-community/playwright-go v0.5700.1
github.com/rivo/tview v0.42.0
github.com/sugarme/tokenizer v0.3.0
github.com/yalue/onnxruntime_go v1.27.0
github.com/yuin/goldmark v1.4.13
)
require (
github.com/andybalholm/cascadia v1.3.3 // indirect
github.com/deckarep/golang-set/v2 v2.8.0 // indirect
github.com/dustin/go-humanize v1.0.1 // indirect
github.com/ebitengine/oto/v3 v3.4.0 // indirect
github.com/ebitengine/purego v0.9.1 // indirect
github.com/emirpasic/gods v1.18.1 // indirect
github.com/gdamore/encoding v1.0.1 // indirect
github.com/go-jose/go-jose/v3 v3.0.4 // indirect
github.com/go-stack/stack v1.8.1 // indirect
github.com/google/uuid v1.6.0 // indirect
github.com/hajimehoshi/go-mp3 v0.3.4 // indirect
github.com/hajimehoshi/oto/v2 v2.3.1 // indirect
github.com/lucasb-eyer/go-colorful v1.3.0 // indirect
github.com/mattn/go-isatty v0.0.20 // indirect
github.com/mitchellh/colorstring v0.0.0-20190213212951-d06e56a500db // indirect
github.com/ncruces/go-strftime v1.0.0 // indirect
github.com/patrickmn/go-cache v2.1.0+incompatible // indirect
github.com/pkg/errors v0.9.1 // indirect
github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec // indirect
github.com/rivo/uniseg v0.4.7 // indirect
github.com/schollz/progressbar/v2 v2.15.0 // indirect
github.com/sugarme/regexpset v0.0.0-20200920021344-4d4ec8eaf93c // indirect
golang.org/x/exp v0.0.0-20251209150349-8475f28825e9 // indirect
golang.org/x/net v0.48.0 // indirect
golang.org/x/sys v0.39.0 // indirect

29
go.sum
View File

@@ -10,22 +10,32 @@ github.com/PuerkitoBio/goquery v1.11.0 h1:jZ7pwMQXIITcUXNH83LLk+txlaEy6NVOfTuP43
github.com/PuerkitoBio/goquery v1.11.0/go.mod h1:wQHgxUOU3JGuj3oD/QFfxUdlzW6xPHfqyHre6VMY4DQ=
github.com/andybalholm/cascadia v1.3.3 h1:AG2YHrzJIm4BZ19iwJ/DAua6Btl3IwJX+VI4kktS1LM=
github.com/andybalholm/cascadia v1.3.3/go.mod h1:xNd9bqTn98Ln4DwST8/nG+H0yuB8Hmgu1YHNnWw0GeA=
github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
github.com/davecgh/go-spew v1.1.1 h1:vj9j/u1bqnvCEfJOwUhtlOARqs3+rkHYY13jYWTU97c=
github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
github.com/deckarep/golang-set/v2 v2.8.0 h1:swm0rlPCmdWn9mESxKOjWk8hXSqoxOp+ZlfuyaAdFlQ=
github.com/deckarep/golang-set/v2 v2.8.0/go.mod h1:VAky9rY/yGXJOLEDv3OMci+7wtDpOF4IN+y82NBOac4=
github.com/dustin/go-humanize v1.0.1 h1:GzkhY7T5VNhEkwH0PVJgjz+fX1rhBrR7pRT3mDkpeCY=
github.com/dustin/go-humanize v1.0.1/go.mod h1:Mu1zIs6XwVuF/gI1OepvI0qD18qycQx+mFykh5fBlto=
github.com/ebitengine/oto/v3 v3.4.0 h1:br0PgASsEWaoWn38b2Goe7m1GKFYfNgnsjSd5Gg+/bQ=
github.com/ebitengine/oto/v3 v3.4.0/go.mod h1:IOleLVD0m+CMak3mRVwsYY8vTctQgOM0iiL6S7Ar7eI=
github.com/ebitengine/purego v0.9.1 h1:a/k2f2HQU3Pi399RPW1MOaZyhKJL9w/xFpKAg4q1s0A=
github.com/ebitengine/purego v0.9.1/go.mod h1:iIjxzd6CiRiOG0UyXP+V1+jWqUXVjPKLAI0mRfJZTmQ=
github.com/emirpasic/gods v1.18.1 h1:FXtiHYKDGKCW2KzwZKx0iC0PQmdlorYgdFG9jPXJ1Bc=
github.com/emirpasic/gods v1.18.1/go.mod h1:8tpGGwCnJ5H4r6BWwaV6OrWmMoPhUl5jm/FMNAnJvWQ=
github.com/gdamore/encoding v1.0.1 h1:YzKZckdBL6jVt2Gc+5p82qhrGiqMdG/eNs6Wy0u3Uhw=
github.com/gdamore/encoding v1.0.1/go.mod h1:0Z0cMFinngz9kS1QfMjCP8TY7em3bZYeeklsSDPivEo=
github.com/gdamore/tcell/v2 v2.13.2 h1:5j4srfF8ow3HICOv/61/sOhQtA25qxEB2XR3Q/Bhx2g=
github.com/gdamore/tcell/v2 v2.13.2/go.mod h1:+Wfe208WDdB7INEtCsNrAN6O2m+wsTPk1RAovjaILlo=
github.com/glebarez/go-sqlite v1.22.0 h1:uAcMJhaA6r3LHMTFgP0SifzgXg46yJkgxqyuyec+ruQ=
github.com/glebarez/go-sqlite v1.22.0/go.mod h1:PlBIdHe0+aUEFn+r2/uthrWq4FxbzugL0L8Li6yQJbc=
github.com/go-jose/go-jose/v3 v3.0.4 h1:Wp5HA7bLQcKnf6YYao/4kpRpVMp/yf6+pJKV8WFSaNY=
github.com/go-jose/go-jose/v3 v3.0.4/go.mod h1:5b+7YgP7ZICgJDBdfjZaIt+H/9L9T/YQrVfLAMboGkQ=
github.com/go-sql-driver/mysql v1.8.1 h1:LedoTUt/eveggdHS9qUFC1EFSa8bU2+1pZjSRpvNJ1Y=
github.com/go-sql-driver/mysql v1.8.1/go.mod h1:wEBSXgmK//2ZFJyE+qWnIsVGmvmEKlqwuVSjsCm7DZg=
github.com/go-stack/stack v1.8.1 h1:ntEHSVwIt7PNXNpgPmVfMrNhLtgjlmnZha2kOpuRiDw=
github.com/go-stack/stack v1.8.1/go.mod h1:dcoOX6HbPZSZptuspn9bctJ+N/CnF5gGygcUP3XYfe4=
github.com/google/go-cmp v0.5.9/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
github.com/google/go-cmp v0.6.0/go.mod h1:17dUlkBOakJ0+DkrSSNjCkIjxS6bF9zb3elmeNGIjoY=
github.com/google/pprof v0.0.0-20250317173921-a4b03ec1a45e h1:ijClszYn+mADRFY17kjQEVQ1XRhq2/JR1M3sGqeJoxs=
github.com/google/pprof v0.0.0-20250317173921-a4b03ec1a45e/go.mod h1:boTsfXsheKC2y+lKOCMpSfarhxDeIzfZG1jqGcPl3cA=
@@ -53,12 +63,18 @@ github.com/mattn/go-isatty v0.0.20 h1:xfD0iDuEKnDkl03q4limB+vH+GxLEtL/jb4xVJSWWE
github.com/mattn/go-isatty v0.0.20/go.mod h1:W+V8PltTTMOvKvAeJH7IuucS94S2C6jfK/D7dTCTo3Y=
github.com/mattn/go-sqlite3 v1.14.22 h1:2gZY6PC6kBnID23Tichd1K+Z0oS6nE/XwU+Vz/5o4kU=
github.com/mattn/go-sqlite3 v1.14.22/go.mod h1:Uh1q+B4BYcTPb+yiD3kU8Ct7aC0hY9fxUwlHK0RXw+Y=
github.com/mitchellh/colorstring v0.0.0-20190213212951-d06e56a500db h1:62I3jR2EmQ4l5rM/4FEfDWcRD+abF5XlKShorW5LRoQ=
github.com/mitchellh/colorstring v0.0.0-20190213212951-d06e56a500db/go.mod h1:l0dey0ia/Uv7NcFFVbCLtqEBQbrT4OCwCSKTEv6enCw=
github.com/ncruces/go-strftime v1.0.0 h1:HMFp8mLCTPp341M/ZnA4qaf7ZlsbTc+miZjCLOFAw7w=
github.com/ncruces/go-strftime v1.0.0/go.mod h1:Fwc5htZGVVkseilnfgOVb9mKy6w1naJmn9CehxcKcls=
github.com/neurosnap/sentences v1.1.2 h1:iphYOzx/XckXeBiLIUBkPu2EKMJ+6jDbz/sLJZ7ZoUw=
github.com/neurosnap/sentences v1.1.2/go.mod h1:/pwU4E9XNL21ygMIkOIllv/SMy2ujHwpf8GQPu1YPbQ=
github.com/patrickmn/go-cache v2.1.0+incompatible h1:HRMgzkcYKYpi3C8ajMPV8OFXaaRUnok+kx1WdO15EQc=
github.com/patrickmn/go-cache v2.1.0+incompatible/go.mod h1:3Qf8kWWT7OJRJbdiICTKqZju1ZixQ/KpMGzzAfe6+WQ=
github.com/pkg/errors v0.9.1 h1:FEBLx1zS214owpjy7qsBeixbURkuhQAwrK5UwLGTwt4=
github.com/pkg/errors v0.9.1/go.mod h1:bwawxfHBFNV+L2hUp1rHADufV3IMtnDRdf1r5NINEl0=
github.com/playwright-community/playwright-go v0.5700.1 h1:PNFb1byWqrTT720rEO0JL88C6Ju0EmUnR5deFLvtP/U=
github.com/playwright-community/playwright-go v0.5700.1/go.mod h1:MlSn1dZrx8rszbCxY6x3qK89ZesJUYVx21B2JnkoNF0=
github.com/pmezard/go-difflib v1.0.0 h1:4DBwDE0NGyQoBHbLQYPwSUPoCMWR5BEzIk/f1lZbAQM=
github.com/pmezard/go-difflib v1.0.0/go.mod h1:iKH77koFhYxTK1pcRnkKkqfTogsbg7gZNVY4sRDYZ/4=
github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec h1:W09IVJc94icq4NjY3clb7Lk8O1qJ8BdBEF8z0ibU0rE=
@@ -67,8 +83,19 @@ github.com/rivo/tview v0.42.0 h1:b/ftp+RxtDsHSaynXTbJb+/n/BxDEi+W3UfF5jILK6c=
github.com/rivo/tview v0.42.0/go.mod h1:cSfIYfhpSGCjp3r/ECJb+GKS7cGJnqV8vfjQPwoXyfY=
github.com/rivo/uniseg v0.4.7 h1:WUdvkW8uEhrYfLC4ZzdpI2ztxP1I582+49Oc5Mq64VQ=
github.com/rivo/uniseg v0.4.7/go.mod h1:FN3SvrM+Zdj16jyLfmOkMNblXMcoc8DfTHruCPUcx88=
github.com/schollz/progressbar/v2 v2.15.0 h1:dVzHQ8fHRmtPjD3K10jT3Qgn/+H+92jhPrhmxIJfDz8=
github.com/schollz/progressbar/v2 v2.15.0/go.mod h1:UdPq3prGkfQ7MOzZKlDRpYKcFqEMczbD7YmbPgpzKMI=
github.com/stretchr/objx v0.1.0/go.mod h1:HFkY916IF+rwdDfMAkV7OtwuqBVzrE8GR6GFx+wExME=
github.com/stretchr/testify v1.3.0/go.mod h1:M5WIy9Dh21IEIfnGCwXGc5bZfKNJtfHm1UVUgZn+9EI=
github.com/stretchr/testify v1.7.0/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/h/Wwjteg=
github.com/stretchr/testify v1.10.0 h1:Xv5erBjTwe/5IxqUQTdXv5kgmIvbHo3QQyRwhJsOfJA=
github.com/stretchr/testify v1.10.0/go.mod h1:r2ic/lqez/lEtzL7wO/rwa5dbSLXVDPFyf8C91i36aY=
github.com/sugarme/regexpset v0.0.0-20200920021344-4d4ec8eaf93c h1:pwb4kNSHb4K89ymCaN+5lPH/MwnfSVg4rzGDh4d+iy4=
github.com/sugarme/regexpset v0.0.0-20200920021344-4d4ec8eaf93c/go.mod h1:2gwkXLWbDGUQWeL3RtpCmcY4mzCtU13kb9UsAg9xMaw=
github.com/sugarme/tokenizer v0.3.0 h1:FE8DYbNSz/kSbgEo9l/RjgYHkIJYEdskumitFQBE9FE=
github.com/sugarme/tokenizer v0.3.0/go.mod h1:VJ+DLK5ZEZwzvODOWwY0cw+B1dabTd3nCB5HuFCItCc=
github.com/yalue/onnxruntime_go v1.27.0 h1:c1YSgDNtpf0WGtxj3YeRIb8VC5LmM1J+Ve3uHdteC1U=
github.com/yalue/onnxruntime_go v1.27.0/go.mod h1:b4X26A8pekNb1ACJ58wAXgNKeUCGEAQ9dmACut9Sm/4=
github.com/yuin/goldmark v1.4.13 h1:fVcFKWvrslecOb/tg+Cc05dkeYx540o0FuFt3nUVDoE=
github.com/yuin/goldmark v1.4.13/go.mod h1:6yULJ656Px+3vBD8DxQVa3kxgyrAnzto9xy5taEt/CY=
golang.org/x/crypto v0.0.0-20190308221718-c2843e01d9a2/go.mod h1:djNgcEr1/C05ACkg1iLfiJU5Ep61QUkGW8qpdssI0+w=
@@ -152,6 +179,8 @@ golang.org/x/tools v0.21.1-0.20240508182429-e35e4ccd0d2d/go.mod h1:aiJjzUbINMkxb
golang.org/x/tools v0.40.0 h1:yLkxfA+Qnul4cs9QA3KnlFu0lVmd8JJfoq+E41uSutA=
golang.org/x/tools v0.40.0/go.mod h1:Ik/tzLRlbscWpqqMRjyWYDisX8bG13FrdXp3o4Sr9lc=
golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
gopkg.in/yaml.v3 v3.0.0-20200313102051-9f266ea9e77c/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
gopkg.in/yaml.v3 v3.0.1 h1:fxVm/GzAzEWqLHuvctI91KS9hhNmmWOoWu0XTYJS7CA=
gopkg.in/yaml.v3 v3.0.1/go.mod h1:K4uyk7z7BCEPqu6E+C64Yfv1cQ7kz7rIZviUmN+EgEM=
modernc.org/cc/v4 v4.27.1 h1:9W30zRlYrefrDV2JE2O8VDtJ1yPGownxciz5rrbQZis=

View File

@@ -11,12 +11,11 @@ import (
"path"
"path/filepath"
"slices"
"strconv"
"strings"
"time"
"unicode"
"math/rand/v2"
"github.com/rivo/tview"
)
@@ -29,7 +28,6 @@ func startModelColorUpdater() {
go func() {
ticker := time.NewTicker(5 * time.Second)
defer ticker.Stop()
// Initial check
updateCachedModelColor()
for range ticker.C {
@@ -44,7 +42,6 @@ func updateCachedModelColor() {
cachedModelColor = "orange"
return
}
// Check if model is loaded
loaded, err := isModelLoaded(chatBody.Model)
if err != nil {
@@ -68,6 +65,14 @@ func isASCII(s string) bool {
return true
}
func mapToString[V any](m map[string]V) string {
rs := strings.Builder{}
for k, v := range m {
fmt.Fprintf(&rs, "%v: %v\n", k, v)
}
return rs.String()
}
// stripThinkingFromMsg removes thinking blocks from assistant messages.
// Skips user, tool, and system messages as they may contain thinking examples.
func stripThinkingFromMsg(msg *models.RoleMsg) *models.RoleMsg {
@@ -193,7 +198,11 @@ func initSysCards() ([]string, error) {
logger.Warn("empty role", "file", cc.FilePath)
continue
}
sysMap[cc.Role] = cc
if cc.ID == "" {
cc.ID = models.ComputeCardID(cc.Role, cc.FilePath)
}
sysMap[cc.ID] = cc
roleToID[cc.Role] = cc.ID
labels = append(labels, cc.Role)
}
return labels, nil
@@ -211,8 +220,10 @@ func startNewChat(keepSysP bool) {
chatBody.Messages = chatBody.Messages[:2]
textView.SetText(chatToText(chatBody.Messages, cfg.ShowSys))
newChat := &models.Chat{
ID: id + 1,
Name: fmt.Sprintf("%d_%s", id+1, cfg.AssistantRole),
ID: id + 1,
Name: fmt.Sprintf("%d_%s", id+1, cfg.AssistantRole),
CreatedAt: time.Now(),
UpdatedAt: time.Now(),
// chat is written to db when we get first llm response (or any)
// actual chat history (messages) would be parsed then
Msgs: "",
@@ -280,24 +291,25 @@ func listRolesWithUser() []string {
return result
}
func loadImage() {
func loadImage() error {
filepath := defaultImage
cc, ok := sysMap[cfg.AssistantRole]
if ok {
cc := GetCardByRole(cfg.AssistantRole)
if cc != nil {
if strings.HasSuffix(cc.FilePath, ".png") {
filepath = cc.FilePath
}
}
file, err := os.Open(filepath)
if err != nil {
panic(err)
return fmt.Errorf("failed to open image: %w", err)
}
defer file.Close()
img, _, err := image.Decode(file)
if err != nil {
panic(err)
return fmt.Errorf("failed to decode image: %w", err)
}
imgView.SetImage(img)
return nil
}
func strInSlice(s string, sl []string) bool {
@@ -357,7 +369,7 @@ func makeStatusLine() string {
}
// Get model color based on load status for local llama.cpp models
modelColor := getModelColor()
statusLine := fmt.Sprintf(statusLineTempl, boolColors[botRespMode], activeChatName,
statusLine := fmt.Sprintf(statusLineTempl, activeChatName,
boolColors[cfg.ToolUse], modelColor, chatBody.Model, boolColors[cfg.SkipLLMResp],
cfg.CurrentAPI, persona, botPersona)
if cfg.STT_ENABLED {
@@ -370,17 +382,88 @@ func makeStatusLine() string {
roleInject := fmt.Sprintf(" | [%s:-:b]role injection[-:-:-] (alt+7)", boolColors[injectRole])
statusLine += roleInject
}
// context tokens
contextTokens := getContextTokens()
maxCtx := getMaxContextTokens()
if maxCtx == 0 {
maxCtx = 16384
}
if contextTokens > 0 {
contextInfo := fmt.Sprintf(" | context-estim: [orange:-:b]%d/%d[-:-:-]", contextTokens, maxCtx)
statusLine += contextInfo
}
return statusLine + imageInfo + shellModeInfo
}
var letters = []rune("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ")
func randString(n int) string {
b := make([]rune, n)
for i := range b {
b[i] = letters[rand.IntN(len(letters))]
func getContextTokens() int {
if chatBody == nil || chatBody.Messages == nil {
return 0
}
return string(b)
total := 0
messages := chatBody.Messages
for i := range messages {
msg := &messages[i]
if msg.Stats != nil && msg.Stats.Tokens > 0 {
total += msg.Stats.Tokens
} else if msg.GetText() != "" {
total += len(msg.GetText()) / 4
}
}
return total
}
const deepseekContext = 128000
func getMaxContextTokens() int {
if chatBody == nil || chatBody.Model == "" {
return 0
}
modelName := chatBody.Model
switch {
case strings.Contains(cfg.CurrentAPI, "openrouter"):
if orModelsData != nil {
for i := range orModelsData.Data {
m := &orModelsData.Data[i]
if m.ID == modelName {
return m.ContextLength
}
}
}
case strings.Contains(cfg.CurrentAPI, "deepseek"):
return deepseekContext
default:
if localModelsData != nil {
for i := range localModelsData.Data {
m := &localModelsData.Data[i]
if m.ID == modelName {
for _, arg := range m.Status.Args {
if strings.HasPrefix(arg, "--ctx-size") {
if strings.Contains(arg, "=") {
val := strings.Split(arg, "=")[1]
if n, err := strconv.Atoi(val); err == nil {
return n
}
} else {
idx := -1
for j, a := range m.Status.Args {
if a == "--ctx-size" && j+1 < len(m.Status.Args) {
idx = j + 1
break
}
}
if idx != -1 {
if n, err := strconv.Atoi(m.Status.Args[idx]); err == nil {
return n
}
}
}
}
}
}
}
}
}
return 0
}
// set of roles within card definition and mention in chat history
@@ -390,13 +473,9 @@ func listChatRoles() []string {
if !ok {
return cbc
}
currentCard, ok := sysMap[currentChat.Agent]
if !ok {
// case which won't let to switch roles:
// started new chat (basic_sys or any other), at the start it yet be saved or have chatbody
// if it does not have a card or chars, it'll return an empty slice
// log error
logger.Warn("failed to find current card in sysMap", "agent", currentChat.Agent, "sysMap", sysMap)
currentCard := GetCardByRole(currentChat.Agent)
if currentCard == nil {
logger.Warn("failed to find current card", "agent", currentChat.Agent)
return cbc
}
charset := []string{}
@@ -412,10 +491,7 @@ func listChatRoles() []string {
func deepseekModelValidator() error {
if cfg.CurrentAPI == cfg.DeepSeekChatAPI || cfg.CurrentAPI == cfg.DeepSeekCompletionAPI {
if chatBody.Model != "deepseek-chat" && chatBody.Model != "deepseek-reasoner" {
if err := notifyUser("bad request", "wrong deepseek model name"); err != nil {
logger.Warn("failed ot notify user", "error", err)
return err
}
showToast("bad request", "wrong deepseek model name")
return nil
}
}
@@ -426,12 +502,11 @@ func deepseekModelValidator() error {
func toggleShellMode() {
shellMode = !shellMode
setShellMode(shellMode)
if shellMode {
// Update input placeholder to indicate shell mode
textArea.SetPlaceholder("SHELL MODE: Enter command and press <Esc> to execute")
shellInput.SetLabel(fmt.Sprintf("[%s]$ ", cfg.FilePickerDir))
} else {
// Reset to normal mode
textArea.SetPlaceholder("input is multiline; press <Enter> to start the next line;\npress <Esc> to send the message. Alt+1 to exit shell mode")
textArea.SetPlaceholder("input is multiline; press <Enter> to start the next line;\npress <Esc> to send the message.")
}
updateStatusLine()
}
@@ -443,23 +518,29 @@ func updateFlexLayout() {
}
flex.Clear()
flex.AddItem(textView, 0, 40, false)
flex.AddItem(textArea, 0, 10, false)
if shellMode {
flex.AddItem(shellInput, 0, 10, false)
} else {
flex.AddItem(bottomFlex, 0, 10, true)
}
if positionVisible {
flex.AddItem(statusLineWidget, 0, 2, false)
}
// Keep focus on currently focused widget
focused := app.GetFocus()
if focused == textView {
switch {
case focused == textView:
app.SetFocus(textView)
} else {
case shellMode:
app.SetFocus(shellInput)
default:
app.SetFocus(textArea)
}
}
func executeCommandAndDisplay(cmdText string) {
// Parse the command (split by spaces, but handle quoted arguments)
cmdParts := parseCommand(cmdText)
if len(cmdParts) == 0 {
cmdText = strings.TrimSpace(cmdText)
if cmdText == "" {
fmt.Fprintf(textView, "\n[red]Error: No command provided[-:-:-]\n")
if scrollToEndEnabled {
textView.ScrollToEnd()
@@ -467,17 +548,63 @@ func executeCommandAndDisplay(cmdText string) {
colorText()
return
}
command := cmdParts[0]
args := []string{}
if len(cmdParts) > 1 {
args = cmdParts[1:]
workingDir := cfg.FilePickerDir
// Handle cd command specially to update working directory
if strings.HasPrefix(cmdText, "cd ") {
newDir := strings.TrimPrefix(cmdText, "cd ")
newDir = strings.TrimSpace(newDir)
// Handle cd ~ or cdHOME
if strings.HasPrefix(newDir, "~") {
home := os.Getenv("HOME")
newDir = strings.Replace(newDir, "~", home, 1)
}
// Check if directory exists
if _, err := os.Stat(newDir); err == nil {
workingDir = newDir
cfg.FilePickerDir = workingDir
// Update shell input label with new directory
shellInput.SetLabel(fmt.Sprintf("[%s]$ ", cfg.FilePickerDir))
outputContent := workingDir
// Add the command being executed to the chat
fmt.Fprintf(textView, "\n[-:-:b](%d) <%s>: [-:-:-]\n$ %s\n",
len(chatBody.Messages), cfg.ToolRole, cmdText)
fmt.Fprintf(textView, "%s\n", outputContent)
combinedMsg := models.RoleMsg{
Role: cfg.ToolRole,
Content: "$ " + cmdText + "\n\n" + outputContent,
}
chatBody.Messages = append(chatBody.Messages, combinedMsg)
if scrollToEndEnabled {
textView.ScrollToEnd()
}
colorText()
return
} else {
outputContent := "cd: " + newDir + ": No such file or directory"
fmt.Fprintf(textView, "\n[-:-:b](%d) <%s>: [-:-:-]\n$ %s\n",
len(chatBody.Messages), cfg.ToolRole, cmdText)
fmt.Fprintf(textView, "[red]%s[-:-:-]\n", outputContent)
combinedMsg := models.RoleMsg{
Role: cfg.ToolRole,
Content: "$ " + cmdText + "\n\n" + outputContent,
}
chatBody.Messages = append(chatBody.Messages, combinedMsg)
if scrollToEndEnabled {
textView.ScrollToEnd()
}
colorText()
return
}
}
// Create the command execution
cmd := exec.Command(command, args...)
// Use /bin/sh to support pipes, redirects, etc.
cmd := exec.Command("/bin/sh", "-c", cmdText)
cmd.Dir = workingDir
// Execute the command and get output
output, err := cmd.CombinedOutput()
// Add the command being executed to the chat
fmt.Fprintf(textView, "\n[yellow]$ %s[-:-:-]\n", cmdText)
fmt.Fprintf(textView, "\n[-:-:b](%d) <%s>: [-:-:-]\n$ %s\n",
len(chatBody.Messages), cfg.ToolRole, cmdText)
var outputContent string
if err != nil {
// Include both output and error
@@ -514,42 +641,11 @@ func executeCommandAndDisplay(cmdText string) {
textView.ScrollToEnd()
}
colorText()
}
// parseCommand splits command string handling quotes properly
func parseCommand(cmd string) []string {
var args []string
var current string
var inQuotes bool
var quoteChar rune
for _, r := range cmd {
switch r {
case '"', '\'':
if inQuotes {
if r == quoteChar {
inQuotes = false
} else {
current += string(r)
}
} else {
inQuotes = true
quoteChar = r
}
case ' ', '\t':
if inQuotes {
current += string(r)
} else if current != "" {
args = append(args, current)
current = ""
}
default:
current += string(r)
}
// Add command to history (avoid duplicates at the end)
if len(shellHistory) == 0 || shellHistory[len(shellHistory)-1] != cmdText {
shellHistory = append(shellHistory, cmdText)
}
if current != "" {
args = append(args, current)
}
return args
shellHistoryPos = -1
}
// == search ==
@@ -595,9 +691,7 @@ func performSearch(term string) {
searchResults = nil
searchResultLengths = nil
notification := "Pattern not found: " + term
if err := notifyUser("search", notification); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("search", notification)
return
}
// Store the formatted text positions and lengths for accurate highlighting
@@ -630,9 +724,7 @@ func highlightCurrentMatch() {
textView.Highlight(currentRegion).ScrollToHighlight()
// Send notification about which match we're at
notification := fmt.Sprintf("Match %d of %d", searchIndex+1, len(searchResults))
if err := notifyUser("search", notification); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("search", notification)
}
// showSearchBar shows the search input field as an overlay
@@ -722,9 +814,7 @@ func addRegionTags(text string, positions []int, lengths []int, currentIdx int,
// searchNext finds the next occurrence of the search term
func searchNext() {
if len(searchResults) == 0 {
if err := notifyUser("search", "No search results to navigate"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("search", "No search results to navigate")
return
}
searchIndex = (searchIndex + 1) % len(searchResults)
@@ -734,9 +824,7 @@ func searchNext() {
// searchPrev finds the previous occurrence of the search term
func searchPrev() {
if len(searchResults) == 0 {
if err := notifyUser("search", "No search results to navigate"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("search", "No search results to navigate")
return
}
if searchIndex == 0 {
@@ -791,3 +879,91 @@ func scanFiles(dir, filter string) []string {
scanRecursive(dir, 0, "")
return files
}
// models logic that is too complex for models package
func MsgToText(i int, m *models.RoleMsg) string {
var contentStr string
var imageIndicators []string
if !m.HasContentParts {
contentStr = m.Content
} else {
var textParts []string
for _, part := range m.ContentParts {
switch p := part.(type) {
case models.TextContentPart:
if p.Type == "text" {
textParts = append(textParts, p.Text)
}
case models.ImageContentPart:
displayPath := p.Path
if displayPath == "" {
displayPath = "image"
} else {
displayPath = extractDisplayPath(displayPath, cfg.FilePickerDir)
}
imageIndicators = append(imageIndicators, fmt.Sprintf("[orange::i][image: %s][-:-:-]", displayPath))
case map[string]any:
if partType, exists := p["type"]; exists {
switch partType {
case "text":
if textVal, textExists := p["text"]; textExists {
if textStr, isStr := textVal.(string); isStr {
textParts = append(textParts, textStr)
}
}
case "image_url":
var displayPath string
if pathVal, pathExists := p["path"]; pathExists {
if pathStr, isStr := pathVal.(string); isStr && pathStr != "" {
displayPath = extractDisplayPath(pathStr, cfg.FilePickerDir)
}
}
if displayPath == "" {
displayPath = "image"
}
imageIndicators = append(imageIndicators, fmt.Sprintf("[orange::i][image: %s][-:-:-]", displayPath))
}
}
}
}
contentStr = strings.Join(textParts, " ") + " "
}
contentStr, _ = strings.CutPrefix(contentStr, m.Role+":")
icon := fmt.Sprintf("(%d) <%s>: ", i, m.Role)
var finalContent strings.Builder
if len(imageIndicators) > 0 {
for _, indicator := range imageIndicators {
finalContent.WriteString(indicator)
finalContent.WriteString("\n")
}
}
finalContent.WriteString(contentStr)
if m.Stats != nil {
fmt.Fprintf(&finalContent, "\n[gray::i][%d tok, %.1fs, %.1f t/s][-:-:-]", m.Stats.Tokens, m.Stats.Duration, m.Stats.TokensPerSec)
}
textMsg := fmt.Sprintf("[-:-:b]%s[-:-:-]\n%s\n", icon, finalContent.String())
return strings.ReplaceAll(textMsg, "\n\n", "\n")
}
// extractDisplayPath returns a path suitable for display, potentially relative to imageBaseDir
func extractDisplayPath(p, bp string) string {
if p == "" {
return ""
}
// If base directory is set, try to make path relative to it
if bp != "" {
if rel, err := filepath.Rel(bp, p); err == nil {
// Check if relative path doesn't start with ".." (meaning it's within base dir)
// If it starts with "..", we might still want to show it as relative
// but for now we show full path if it goes outside base dir
if !strings.HasPrefix(rel, "..") {
p = rel
}
}
}
// Truncate long paths to last 60 characters if needed
if len(p) > 60 {
return "..." + p[len(p)-60:]
}
return p
}

172
llm.go
View File

@@ -3,7 +3,6 @@ package main
import (
"bytes"
"encoding/json"
"errors"
"gf-lt/models"
"io"
"strings"
@@ -14,8 +13,8 @@ var lastImg string // for ctrl+j
// containsToolSysMsg checks if the toolSysMsg already exists in the chat body
func containsToolSysMsg() bool {
for _, msg := range chatBody.Messages {
if msg.Role == cfg.ToolRole && msg.Content == toolSysMsg {
for i := range chatBody.Messages {
if chatBody.Messages[i].Role == cfg.ToolRole && chatBody.Messages[i].Content == toolSysMsg {
return true
}
}
@@ -119,25 +118,22 @@ func (lcp LCPCompletion) FormMsg(msg, role string, resume bool) (io.Reader, erro
logger.Debug("formmsg lcpcompletion", "link", cfg.CurrentAPI)
localImageAttachmentPath := imageAttachmentPath
var multimodalData []string
if localImageAttachmentPath != "" {
imageURL, err := models.CreateImageURLFromPath(localImageAttachmentPath)
if err != nil {
logger.Error("failed to create image URL from path for completion",
"error", err, "path", localImageAttachmentPath)
return nil, err
}
// Extract base64 part from data URL (e.g., "data:image/jpeg;base64,...")
parts := strings.SplitN(imageURL, ",", 2)
if len(parts) == 2 {
multimodalData = append(multimodalData, parts[1])
} else {
logger.Error("invalid image data URL format", "url", imageURL)
return nil, errors.New("invalid image data URL format")
}
imageAttachmentPath = "" // Clear the attachment after use
}
if msg != "" { // otherwise let the bot to continue
newMsg := models.RoleMsg{Role: role, Content: msg}
var newMsg models.RoleMsg
if localImageAttachmentPath != "" {
newMsg = models.NewMultimodalMsg(role, []any{})
newMsg.AddTextPart(msg)
imageURL, err := models.CreateImageURLFromPath(localImageAttachmentPath)
if err != nil {
logger.Error("failed to create image URL from path for completion",
"error", err, "path", localImageAttachmentPath)
return nil, err
}
newMsg.AddImagePart(imageURL, localImageAttachmentPath)
imageAttachmentPath = "" // Clear the attachment after use
} else { // not a multimodal msg or image passed in tool call
newMsg = models.RoleMsg{Role: role, Content: msg}
}
newMsg = *processMessageTag(&newMsg)
chatBody.Messages = append(chatBody.Messages, newMsg)
}
@@ -146,22 +142,40 @@ func (lcp LCPCompletion) FormMsg(msg, role string, resume bool) (io.Reader, erro
chatBody.Messages = append(chatBody.Messages, models.RoleMsg{Role: cfg.ToolRole, Content: toolSysMsg})
}
filteredMessages, botPersona := filterMessagesForCurrentCharacter(chatBody.Messages)
// Build prompt and extract images inline as we process each message
messages := make([]string, len(filteredMessages))
for i, m := range filteredMessages {
messages[i] = stripThinkingFromMsg(&m).ToPrompt()
for i := range filteredMessages {
m := stripThinkingFromMsg(&filteredMessages[i])
messages[i] = m.ToPrompt()
// Extract images from this message and add marker inline
if len(m.ContentParts) > 0 {
for _, part := range m.ContentParts {
var imgURL string
// Check for struct type
if imgPart, ok := part.(models.ImageContentPart); ok {
imgURL = imgPart.ImageURL.URL
} else if partMap, ok := part.(map[string]any); ok {
// Check for map type (from JSON unmarshaling)
if partType, exists := partMap["type"]; exists && partType == "image_url" {
if imgURLMap, ok := partMap["image_url"].(map[string]any); ok {
if url, ok := imgURLMap["url"].(string); ok {
imgURL = url
}
}
}
}
if imgURL != "" {
// Extract base64 part from data URL (e.g., "data:image/jpeg;base64,...")
parts := strings.SplitN(imgURL, ",", 2)
if len(parts) == 2 {
multimodalData = append(multimodalData, parts[1])
messages[i] += " <__media__>"
}
}
}
}
}
prompt := strings.Join(messages, "\n")
// Add multimodal media markers to the prompt text when multimodal data is present
// This is required by llama.cpp multimodal models so they know where to insert media
if len(multimodalData) > 0 {
// Add a media marker for each item in the multimodal data
var sb strings.Builder
sb.WriteString(prompt)
for range multimodalData {
sb.WriteString(" <__media__>") // llama.cpp default multimodal marker
}
prompt = sb.String()
}
// needs to be after <__media__> if there are images
if !resume {
botMsgStart := "\n" + botPersona + ":\n"
@@ -210,11 +224,9 @@ func (op LCPChat) ParseChunk(data []byte) (*models.TextChunk, error) {
logger.Error("failed to decode", "error", err, "line", string(data))
return nil, err
}
// Handle multiple choices safely
if len(llmchunk.Choices) == 0 {
logger.Warn("LCPChat ParseChunk: no choices in response", "data", string(data))
return &models.TextChunk{Finished: true}, nil
logger.Warn("LCPChat empty chunk choices", "raw_data", string(data), "chunk", llmchunk)
return &models.TextChunk{}, nil
}
lastChoice := llmchunk.Choices[len(llmchunk.Choices)-1]
resp := &models.TextChunk{
@@ -289,14 +301,23 @@ func (op LCPChat) FormMsg(msg, role string, resume bool) (io.Reader, error) {
Model: chatBody.Model,
Stream: chatBody.Stream,
}
for i, msg := range filteredMessages {
strippedMsg := *stripThinkingFromMsg(&msg)
if strippedMsg.Role == cfg.UserRole {
for i := range filteredMessages {
strippedMsg := *stripThinkingFromMsg(&filteredMessages[i])
switch strippedMsg.Role {
case cfg.UserRole:
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "user"
} else {
case cfg.AssistantRole:
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "assistant"
case cfg.ToolRole:
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "tool"
default:
bodyCopy.Messages[i] = strippedMsg
}
// Clear ToolCalls - they're stored in chat history for display but not sent to LLM
// bodyCopy.Messages[i].ToolCall = nil
}
// Clean null/empty messages to prevent API issues
bodyCopy.Messages = consolidateAssistantMessages(bodyCopy.Messages)
@@ -326,6 +347,10 @@ func (ds DeepSeekerCompletion) ParseChunk(data []byte) (*models.TextChunk, error
logger.Error("failed to decode", "error", err, "line", string(data))
return nil, err
}
if len(llmchunk.Choices) == 0 {
logger.Warn("empty chunk choices", "raw_data", string(data), "chunk", llmchunk)
return &models.TextChunk{}, nil
}
resp := &models.TextChunk{
Chunk: llmchunk.Choices[0].Text,
}
@@ -358,8 +383,8 @@ func (ds DeepSeekerCompletion) FormMsg(msg, role string, resume bool) (io.Reader
}
filteredMessages, botPersona := filterMessagesForCurrentCharacter(chatBody.Messages)
messages := make([]string, len(filteredMessages))
for i, m := range filteredMessages {
messages[i] = stripThinkingFromMsg(&m).ToPrompt()
for i := range filteredMessages {
messages[i] = stripThinkingFromMsg(&filteredMessages[i]).ToPrompt()
}
prompt := strings.Join(messages, "\n")
// strings builder?
@@ -391,6 +416,10 @@ func (ds DeepSeekerChat) ParseChunk(data []byte) (*models.TextChunk, error) {
return nil, err
}
resp := &models.TextChunk{}
if len(llmchunk.Choices) == 0 {
logger.Warn("empty chunk choices", "raw_data", string(data), "chunk", llmchunk)
return resp, nil
}
if llmchunk.Choices[0].FinishReason != "" {
if llmchunk.Choices[0].Delta.Content != "" {
logger.Error("text inside of finish llmchunk", "chunk", llmchunk)
@@ -429,14 +458,27 @@ func (ds DeepSeekerChat) FormMsg(msg, role string, resume bool) (io.Reader, erro
Model: chatBody.Model,
Stream: chatBody.Stream,
}
for i, msg := range filteredMessages {
strippedMsg := *stripThinkingFromMsg(&msg)
if strippedMsg.Role == cfg.UserRole || i == 1 {
for i := range filteredMessages {
strippedMsg := *stripThinkingFromMsg(&filteredMessages[i])
switch strippedMsg.Role {
case cfg.UserRole:
if i == 1 {
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "user"
} else {
bodyCopy.Messages[i] = strippedMsg
}
case cfg.AssistantRole:
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "user"
} else {
bodyCopy.Messages[i].Role = "assistant"
case cfg.ToolRole:
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "tool"
default:
bodyCopy.Messages[i] = strippedMsg
}
// Clear ToolCalls - they're stored in chat history for display but not sent to LLM
// bodyCopy.Messages[i].ToolCall = nil
}
// Clean null/empty messages to prevent API issues
bodyCopy.Messages = consolidateAssistantMessages(bodyCopy.Messages)
@@ -460,6 +502,10 @@ func (or OpenRouterCompletion) ParseChunk(data []byte) (*models.TextChunk, error
logger.Error("failed to decode", "error", err, "line", string(data))
return nil, err
}
if len(llmchunk.Choices) == 0 {
logger.Warn("empty chunk choices", "raw_data", string(data), "chunk", llmchunk)
return &models.TextChunk{}, nil
}
resp := &models.TextChunk{
Chunk: llmchunk.Choices[len(llmchunk.Choices)-1].Text,
}
@@ -489,8 +535,8 @@ func (or OpenRouterCompletion) FormMsg(msg, role string, resume bool) (io.Reader
}
filteredMessages, botPersona := filterMessagesForCurrentCharacter(chatBody.Messages)
messages := make([]string, len(filteredMessages))
for i, m := range filteredMessages {
messages[i] = stripThinkingFromMsg(&m).ToPrompt()
for i := range filteredMessages {
messages[i] = stripThinkingFromMsg(&filteredMessages[i]).ToPrompt()
}
prompt := strings.Join(messages, "\n")
// strings builder?
@@ -522,6 +568,10 @@ func (or OpenRouterChat) ParseChunk(data []byte) (*models.TextChunk, error) {
logger.Error("failed to decode", "error", err, "line", string(data))
return nil, err
}
if len(llmchunk.Choices) == 0 {
logger.Warn("empty chunk choices", "raw_data", string(data), "chunk", llmchunk)
return &models.TextChunk{}, nil
}
lastChoice := llmchunk.Choices[len(llmchunk.Choices)-1]
resp := &models.TextChunk{
Chunk: lastChoice.Delta.Content,
@@ -593,14 +643,24 @@ func (or OpenRouterChat) FormMsg(msg, role string, resume bool) (io.Reader, erro
Model: chatBody.Model,
Stream: chatBody.Stream,
}
for i, msg := range filteredMessages {
strippedMsg := *stripThinkingFromMsg(&msg)
bodyCopy.Messages[i] = strippedMsg
// Standardize role if it's a user role
if bodyCopy.Messages[i].Role == cfg.UserRole {
for i := range filteredMessages {
strippedMsg := *stripThinkingFromMsg(&filteredMessages[i])
switch strippedMsg.Role {
case cfg.UserRole:
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "user"
case cfg.AssistantRole:
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "assistant"
case cfg.ToolRole:
bodyCopy.Messages[i] = strippedMsg
bodyCopy.Messages[i].Role = "tool"
default:
bodyCopy.Messages[i] = strippedMsg
}
// Clear ToolCalls - they're stored in chat history for display but not sent to LLM
// literally deletes data that we need
// bodyCopy.Messages[i].ToolCall = nil
}
// Clean null/empty messages to prevent API issues
bodyCopy.Messages = consolidateAssistantMessages(bodyCopy.Messages)

10
main.go
View File

@@ -13,9 +13,13 @@ var (
injectRole = true
selectedIndex = int(-1)
shellMode = false
thinkingCollapsed = false
statusLineTempl = "help (F12) | [%s:-:b]llm writes[-:-:-] (F6 to interrupt) | chat: [orange:-:b]%s[-:-:-] (F1) | [%s:-:b]tool use[-:-:-] (ctrl+k) | model: [%s:-:b]%s[-:-:-] (ctrl+l) | [%s:-:b]skip LLM resp[-:-:-] (F10)\nAPI: [orange:-:b]%s[-:-:-] (ctrl+v) | writing as: [orange:-:b]%s[-:-:-] (ctrl+q) | bot will write as [orange:-:b]%s[-:-:-] (ctrl+x)"
focusSwitcher = map[tview.Primitive]tview.Primitive{}
shellHistory []string
shellHistoryPos int = -1
thinkingCollapsed = false
toolCollapsed = true
statusLineTempl = "help (F12) | chat: [orange:-:b]%s[-:-:-] (F1) | [%s:-:b]tool use[-:-:-] (ctrl+k) | model: [%s:-:b]%s[-:-:-] (ctrl+l) | [%s:-:b]skip LLM resp[-:-:-] (F10) | API: [orange:-:b]%s[-:-:-] (ctrl+v)\nwriting as: [orange:-:b]%s[-:-:-] (ctrl+q) | bot will write as [orange:-:b]%s[-:-:-] (ctrl+x)"
focusSwitcher = map[tview.Primitive]tview.Primitive{}
app *tview.Application
)
func main() {

View File

@@ -1,42 +0,0 @@
package main
import (
"fmt"
"gf-lt/config"
"gf-lt/models"
"strings"
"testing"
)
func TestRemoveThinking(t *testing.T) {
cases := []struct {
cb *models.ChatBody
toolMsgs uint8
}{
{cb: &models.ChatBody{
Stream: true,
Messages: []models.RoleMsg{
{Role: "tool", Content: "should be ommited"},
{Role: "system", Content: "should stay"},
{Role: "user", Content: "hello, how are you?"},
{Role: "assistant", Content: "Oh, hi. <think>I should thank user and continue the conversation</think> I am geat, thank you! How are you?"},
},
},
toolMsgs: uint8(1),
},
}
for i, tc := range cases {
t.Run(fmt.Sprintf("run_%d", i), func(t *testing.T) {
cfg = &config.Config{ToolRole: "tool"} // Initialize cfg.ToolRole for test
mNum := len(tc.cb.Messages)
removeThinking(tc.cb)
if len(tc.cb.Messages) != mNum-int(tc.toolMsgs) {
t.Errorf("failed to delete tools msg %v; expected %d, got %d", tc.cb.Messages, mNum-int(tc.toolMsgs), len(tc.cb.Messages))
}
for _, msg := range tc.cb.Messages {
if strings.Contains(msg.Content, "<think>") {
t.Errorf("msg contains think tag; msg: %s\n", msg.Content)
}
}
}) }
}

View File

@@ -1,6 +1,10 @@
package models
import "strings"
import (
"crypto/md5"
"fmt"
"strings"
)
// https://github.com/malfoyslastname/character-card-spec-v2/blob/main/spec_v2.md
// what a bloat; trim to Role->Msg pair and first msg
@@ -31,6 +35,7 @@ func (c *CharCardSpec) Simplify(userName, fpath string) *CharCard {
fm := strings.ReplaceAll(strings.ReplaceAll(c.FirstMes, "{{char}}", c.Name), "{{user}}", userName)
sysPr := strings.ReplaceAll(strings.ReplaceAll(c.Description, "{{char}}", c.Name), "{{user}}", userName)
return &CharCard{
ID: ComputeCardID(c.Name, fpath),
SysPrompt: sysPr,
FirstMsg: fm,
Role: c.Name,
@@ -39,7 +44,12 @@ func (c *CharCardSpec) Simplify(userName, fpath string) *CharCard {
}
}
func ComputeCardID(role, filePath string) string {
return fmt.Sprintf("%x", md5.Sum([]byte(role+filePath)))
}
type CharCard struct {
ID string `json:"id"`
SysPrompt string `json:"sys_prompt"`
FirstMsg string `json:"first_msg"`
Role string `json:"role"`

13
models/consts.go Normal file
View File

@@ -0,0 +1,13 @@
package models
const (
LoadedMark = "(loaded) "
ToolRespMultyType = "multimodel_content"
)
type APIType int
const (
APITypeChat APIType = iota
APITypeCompletion
)

View File

@@ -5,28 +5,21 @@ import (
"encoding/json"
"fmt"
"os"
"path/filepath"
"strings"
)
var (
// imageBaseDir is the base directory for displaying image paths.
// If set, image paths will be shown relative to this directory.
imageBaseDir = ""
)
// SetImageBaseDir sets the base directory for displaying image paths.
// If dir is empty, full paths will be shown.
func SetImageBaseDir(dir string) {
imageBaseDir = dir
}
type FuncCall struct {
ID string `json:"id,omitempty"`
Name string `json:"name"`
Args map[string]string `json:"args"`
}
type ToolCall struct {
ID string `json:"id,omitempty"`
Name string `json:"name"`
Args string `json:"arguments"`
}
type LLMResp struct {
Choices []struct {
FinishReason string `json:"finish_reason"`
@@ -108,40 +101,56 @@ type RoleMsg struct {
Role string `json:"role"`
Content string `json:"-"`
ContentParts []any `json:"-"`
ToolCallID string `json:"tool_call_id,omitempty"` // For tool response messages
ToolCallID string `json:"tool_call_id,omitempty"` // For tool response messages
ToolCall *ToolCall `json:"tool_call,omitempty"` // For assistant messages with tool calls
IsShellCommand bool `json:"is_shell_command,omitempty"` // True for shell command outputs (always shown)
KnownTo []string `json:"known_to,omitempty"`
Stats *ResponseStats `json:"stats"`
hasContentParts bool // Flag to indicate which content type to marshal
HasContentParts bool // Flag to indicate which content type to marshal
}
// MarshalJSON implements custom JSON marshaling for RoleMsg
func (m *RoleMsg) MarshalJSON() ([]byte, error) {
if m.hasContentParts {
//
//nolint:gocritic
func (m RoleMsg) MarshalJSON() ([]byte, error) {
if m.HasContentParts {
// Use structured content format
aux := struct {
Role string `json:"role"`
Content []any `json:"content"`
ToolCallID string `json:"tool_call_id,omitempty"`
KnownTo []string `json:"known_to,omitempty"`
Role string `json:"role"`
Content []any `json:"content"`
ToolCallID string `json:"tool_call_id,omitempty"`
ToolCall *ToolCall `json:"tool_call,omitempty"`
IsShellCommand bool `json:"is_shell_command,omitempty"`
KnownTo []string `json:"known_to,omitempty"`
Stats *ResponseStats `json:"stats,omitempty"`
}{
Role: m.Role,
Content: m.ContentParts,
ToolCallID: m.ToolCallID,
KnownTo: m.KnownTo,
Role: m.Role,
Content: m.ContentParts,
ToolCallID: m.ToolCallID,
ToolCall: m.ToolCall,
IsShellCommand: m.IsShellCommand,
KnownTo: m.KnownTo,
Stats: m.Stats,
}
return json.Marshal(aux)
} else {
// Use simple content format
aux := struct {
Role string `json:"role"`
Content string `json:"content"`
ToolCallID string `json:"tool_call_id,omitempty"`
KnownTo []string `json:"known_to,omitempty"`
Role string `json:"role"`
Content string `json:"content"`
ToolCallID string `json:"tool_call_id,omitempty"`
ToolCall *ToolCall `json:"tool_call,omitempty"`
IsShellCommand bool `json:"is_shell_command,omitempty"`
KnownTo []string `json:"known_to,omitempty"`
Stats *ResponseStats `json:"stats,omitempty"`
}{
Role: m.Role,
Content: m.Content,
ToolCallID: m.ToolCallID,
KnownTo: m.KnownTo,
Role: m.Role,
Content: m.Content,
ToolCallID: m.ToolCallID,
ToolCall: m.ToolCall,
IsShellCommand: m.IsShellCommand,
KnownTo: m.KnownTo,
Stats: m.Stats,
}
return json.Marshal(aux)
}
@@ -151,26 +160,35 @@ func (m *RoleMsg) MarshalJSON() ([]byte, error) {
func (m *RoleMsg) UnmarshalJSON(data []byte) error {
// First, try to unmarshal as structured content format
var structured struct {
Role string `json:"role"`
Content []any `json:"content"`
ToolCallID string `json:"tool_call_id,omitempty"`
KnownTo []string `json:"known_to,omitempty"`
Role string `json:"role"`
Content []any `json:"content"`
ToolCallID string `json:"tool_call_id,omitempty"`
ToolCall *ToolCall `json:"tool_call,omitempty"`
IsShellCommand bool `json:"is_shell_command,omitempty"`
KnownTo []string `json:"known_to,omitempty"`
Stats *ResponseStats `json:"stats,omitempty"`
}
if err := json.Unmarshal(data, &structured); err == nil && len(structured.Content) > 0 {
m.Role = structured.Role
m.ContentParts = structured.Content
m.ToolCallID = structured.ToolCallID
m.ToolCall = structured.ToolCall
m.IsShellCommand = structured.IsShellCommand
m.KnownTo = structured.KnownTo
m.hasContentParts = true
m.Stats = structured.Stats
m.HasContentParts = true
return nil
}
// Otherwise, unmarshal as simple content format
var simple struct {
Role string `json:"role"`
Content string `json:"content"`
ToolCallID string `json:"tool_call_id,omitempty"`
KnownTo []string `json:"known_to,omitempty"`
Role string `json:"role"`
Content string `json:"content"`
ToolCallID string `json:"tool_call_id,omitempty"`
ToolCall *ToolCall `json:"tool_call,omitempty"`
IsShellCommand bool `json:"is_shell_command,omitempty"`
KnownTo []string `json:"known_to,omitempty"`
Stats *ResponseStats `json:"stats,omitempty"`
}
if err := json.Unmarshal(data, &simple); err != nil {
return err
@@ -178,78 +196,17 @@ func (m *RoleMsg) UnmarshalJSON(data []byte) error {
m.Role = simple.Role
m.Content = simple.Content
m.ToolCallID = simple.ToolCallID
m.ToolCall = simple.ToolCall
m.IsShellCommand = simple.IsShellCommand
m.KnownTo = simple.KnownTo
m.hasContentParts = false
m.Stats = simple.Stats
m.HasContentParts = false
return nil
}
func (m *RoleMsg) ToText(i int) string {
var contentStr string
var imageIndicators []string
if !m.hasContentParts {
contentStr = m.Content
} else {
var textParts []string
for _, part := range m.ContentParts {
switch p := part.(type) {
case TextContentPart:
if p.Type == "text" {
textParts = append(textParts, p.Text)
}
case ImageContentPart:
displayPath := p.Path
if displayPath == "" {
displayPath = "image"
} else {
displayPath = extractDisplayPath(displayPath)
}
imageIndicators = append(imageIndicators, fmt.Sprintf("[orange::i][image: %s][-:-:-]", displayPath))
case map[string]any:
if partType, exists := p["type"]; exists {
switch partType {
case "text":
if textVal, textExists := p["text"]; textExists {
if textStr, isStr := textVal.(string); isStr {
textParts = append(textParts, textStr)
}
}
case "image_url":
var displayPath string
if pathVal, pathExists := p["path"]; pathExists {
if pathStr, isStr := pathVal.(string); isStr && pathStr != "" {
displayPath = extractDisplayPath(pathStr)
}
}
if displayPath == "" {
displayPath = "image"
}
imageIndicators = append(imageIndicators, fmt.Sprintf("[orange::i][image: %s][-:-:-]", displayPath))
}
}
}
}
contentStr = strings.Join(textParts, " ") + " "
}
contentStr, _ = strings.CutPrefix(contentStr, m.Role+":")
icon := fmt.Sprintf("(%d) <%s>: ", i, m.Role)
var finalContent strings.Builder
if len(imageIndicators) > 0 {
for _, indicator := range imageIndicators {
finalContent.WriteString(indicator)
finalContent.WriteString("\n")
}
}
finalContent.WriteString(contentStr)
if m.Stats != nil {
fmt.Fprintf(&finalContent, "\n[gray::i][%d tok, %.1fs, %.1f t/s][-:-:-]", m.Stats.Tokens, m.Stats.Duration, m.Stats.TokensPerSec)
}
textMsg := fmt.Sprintf("[-:-:b]%s[-:-:-]\n%s\n", icon, finalContent.String())
return strings.ReplaceAll(textMsg, "\n\n", "\n")
}
func (m *RoleMsg) ToPrompt() string {
var contentStr string
if !m.hasContentParts {
if !m.HasContentParts {
contentStr = m.Content
} else {
// For structured content, just take the text parts
@@ -282,7 +239,7 @@ func NewRoleMsg(role, content string) RoleMsg {
return RoleMsg{
Role: role,
Content: content,
hasContentParts: false,
HasContentParts: false,
}
}
@@ -291,7 +248,7 @@ func NewMultimodalMsg(role string, contentParts []any) RoleMsg {
return RoleMsg{
Role: role,
ContentParts: contentParts,
hasContentParts: true,
HasContentParts: true,
}
}
@@ -300,7 +257,7 @@ func (m *RoleMsg) HasContent() bool {
if m.Content != "" {
return true
}
if m.hasContentParts && len(m.ContentParts) > 0 {
if m.HasContentParts && len(m.ContentParts) > 0 {
return true
}
return false
@@ -308,7 +265,7 @@ func (m *RoleMsg) HasContent() bool {
// IsContentParts returns true if the message uses structured content parts
func (m *RoleMsg) IsContentParts() bool {
return m.hasContentParts
return m.HasContentParts
}
// GetContentParts returns the content parts of the message
@@ -325,14 +282,16 @@ func (m *RoleMsg) Copy() RoleMsg {
ToolCallID: m.ToolCallID,
KnownTo: m.KnownTo,
Stats: m.Stats,
hasContentParts: m.hasContentParts,
HasContentParts: m.HasContentParts,
ToolCall: m.ToolCall,
IsShellCommand: m.IsShellCommand,
}
}
// GetText returns the text content of the message, handling both
// simple Content and multimodal ContentParts formats.
func (m *RoleMsg) GetText() string {
if !m.hasContentParts {
if !m.HasContentParts {
return m.Content
}
var textParts []string
@@ -361,7 +320,7 @@ func (m *RoleMsg) GetText() string {
// ContentParts (multimodal), it updates the text parts while preserving
// images. If not, it sets the simple Content field.
func (m *RoleMsg) SetText(text string) {
if !m.hasContentParts {
if !m.HasContentParts {
m.Content = text
return
}
@@ -391,14 +350,14 @@ func (m *RoleMsg) SetText(text string) {
// AddTextPart adds a text content part to the message
func (m *RoleMsg) AddTextPart(text string) {
if !m.hasContentParts {
if !m.HasContentParts {
// Convert to content parts format
if m.Content != "" {
m.ContentParts = []any{TextContentPart{Type: "text", Text: m.Content}}
} else {
m.ContentParts = []any{}
}
m.hasContentParts = true
m.HasContentParts = true
}
textPart := TextContentPart{Type: "text", Text: text}
m.ContentParts = append(m.ContentParts, textPart)
@@ -406,14 +365,14 @@ func (m *RoleMsg) AddTextPart(text string) {
// AddImagePart adds an image content part to the message
func (m *RoleMsg) AddImagePart(imageURL, imagePath string) {
if !m.hasContentParts {
if !m.HasContentParts {
// Convert to content parts format
if m.Content != "" {
m.ContentParts = []any{TextContentPart{Type: "text", Text: m.Content}}
} else {
m.ContentParts = []any{}
}
m.hasContentParts = true
m.HasContentParts = true
}
imagePart := ImageContentPart{
Type: "image_url",
@@ -432,7 +391,6 @@ func CreateImageURLFromPath(imagePath string) (string, error) {
if err != nil {
return "", err
}
// Determine the image format based on file extension
var mimeType string
switch {
@@ -449,39 +407,12 @@ func CreateImageURLFromPath(imagePath string) (string, error) {
default:
mimeType = "image/jpeg" // default
}
// Encode to base64
encoded := base64.StdEncoding.EncodeToString(data)
// Create data URL
return fmt.Sprintf("data:%s;base64,%s", mimeType, encoded), nil
}
// extractDisplayPath returns a path suitable for display, potentially relative to imageBaseDir
func extractDisplayPath(p string) string {
if p == "" {
return ""
}
// If base directory is set, try to make path relative to it
if imageBaseDir != "" {
if rel, err := filepath.Rel(imageBaseDir, p); err == nil {
// Check if relative path doesn't start with ".." (meaning it's within base dir)
// If it starts with "..", we might still want to show it as relative
// but for now we show full path if it goes outside base dir
if !strings.HasPrefix(rel, "..") {
p = rel
}
}
}
// Truncate long paths to last 60 characters if needed
if len(p) > 60 {
return "..." + p[len(p)-60:]
}
return p
}
type ChatBody struct {
Model string `json:"model"`
Stream bool `json:"stream"`
@@ -489,16 +420,16 @@ type ChatBody struct {
}
func (cb *ChatBody) Rename(oldname, newname string) {
for i, m := range cb.Messages {
cb.Messages[i].Content = strings.ReplaceAll(m.Content, oldname, newname)
cb.Messages[i].Role = strings.ReplaceAll(m.Role, oldname, newname)
for i := range cb.Messages {
cb.Messages[i].Content = strings.ReplaceAll(cb.Messages[i].Content, oldname, newname)
cb.Messages[i].Role = strings.ReplaceAll(cb.Messages[i].Role, oldname, newname)
}
}
func (cb *ChatBody) ListRoles() []string {
namesMap := make(map[string]struct{})
for _, m := range cb.Messages {
namesMap[m.Role] = struct{}{}
for i := range cb.Messages {
namesMap[cb.Messages[i].Role] = struct{}{}
}
resp := make([]string, len(namesMap))
i := 0
@@ -585,24 +516,6 @@ type OpenAIReq struct {
// ===
// type LLMModels struct {
// Object string `json:"object"`
// Data []struct {
// ID string `json:"id"`
// Object string `json:"object"`
// Created int `json:"created"`
// OwnedBy string `json:"owned_by"`
// Meta struct {
// VocabType int `json:"vocab_type"`
// NVocab int `json:"n_vocab"`
// NCtxTrain int `json:"n_ctx_train"`
// NEmbd int `json:"n_embd"`
// NParams int64 `json:"n_params"`
// Size int64 `json:"size"`
// } `json:"meta"`
// } `json:"data"`
// }
type LlamaCPPReq struct {
Model string `json:"model"`
Stream bool `json:"stream"`
@@ -695,6 +608,20 @@ func (lcp *LCPModels) ListModels() []string {
return resp
}
func (lcp *LCPModels) HasVision(modelID string) bool {
for _, m := range lcp.Data {
if m.ID == modelID {
args := m.Status.Args
for i := 0; i < len(args)-1; i++ {
if args[i] == "--mmproj" {
return true
}
}
}
}
return false
}
type ResponseStats struct {
Tokens int
Duration float64
@@ -708,9 +635,7 @@ type ChatRoundReq struct {
Resume bool
}
type APIType int
const (
APITypeChat APIType = iota
APITypeCompletion
)
type MultimodalToolResp struct {
Type string `json:"type"`
Parts []map[string]string `json:"parts"`
}

View File

@@ -1,161 +0,0 @@
package models
import (
"strings"
"testing"
)
func TestRoleMsgToTextWithImages(t *testing.T) {
tests := []struct {
name string
msg RoleMsg
index int
expected string // substring to check
}{
{
name: "text and image",
index: 0,
msg: func() RoleMsg {
msg := NewMultimodalMsg("user", []interface{}{})
msg.AddTextPart("Look at this picture")
msg.AddImagePart("data:image/jpeg;base64,abc123", "/home/user/Pictures/cat.jpg")
return msg
}(),
expected: "[orange::i][image: /home/user/Pictures/cat.jpg][-:-:-]",
},
{
name: "image only",
index: 1,
msg: func() RoleMsg {
msg := NewMultimodalMsg("user", []interface{}{})
msg.AddImagePart("data:image/png;base64,xyz789", "/tmp/screenshot_20250217_123456.png")
return msg
}(),
expected: "[orange::i][image: /tmp/screenshot_20250217_123456.png][-:-:-]",
},
{
name: "long filename truncated",
index: 2,
msg: func() RoleMsg {
msg := NewMultimodalMsg("user", []interface{}{})
msg.AddTextPart("Check this")
msg.AddImagePart("data:image/jpeg;base64,foo", "/very/long/path/to/a/really_long_filename_that_exceeds_forty_characters.jpg")
return msg
}(),
expected: "[orange::i][image: .../to/a/really_long_filename_that_exceeds_forty_characters.jpg][-:-:-]",
},
{
name: "multiple images",
index: 3,
msg: func() RoleMsg {
msg := NewMultimodalMsg("user", []interface{}{})
msg.AddTextPart("Multiple images")
msg.AddImagePart("data:image/jpeg;base64,a", "/path/img1.jpg")
msg.AddImagePart("data:image/png;base64,b", "/path/img2.png")
return msg
}(),
expected: "[orange::i][image: /path/img1.jpg][-:-:-]\n[orange::i][image: /path/img2.png][-:-:-]",
},
{
name: "old format without path",
index: 4,
msg: RoleMsg{
Role: "user",
hasContentParts: true,
ContentParts: []interface{}{
map[string]interface{}{
"type": "image_url",
"image_url": map[string]interface{}{
"url": "data:image/jpeg;base64,old",
},
},
},
},
expected: "[orange::i][image: image][-:-:-]",
},
{
name: "old format with path",
index: 5,
msg: RoleMsg{
Role: "user",
hasContentParts: true,
ContentParts: []interface{}{
map[string]interface{}{
"type": "image_url",
"path": "/old/path/photo.jpg",
"image_url": map[string]interface{}{
"url": "data:image/jpeg;base64,old",
},
},
},
},
expected: "[orange::i][image: /old/path/photo.jpg][-:-:-]",
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
result := tt.msg.ToText(tt.index)
if !strings.Contains(result, tt.expected) {
t.Errorf("ToText() result does not contain expected indicator\ngot: %s\nwant substring: %s", result, tt.expected)
}
// Ensure the indicator appears before text content
if strings.Contains(tt.expected, "cat.jpg") && strings.Contains(result, "Look at this picture") {
indicatorPos := strings.Index(result, "[orange::i][image: /home/user/Pictures/cat.jpg][-:-:-]")
textPos := strings.Index(result, "Look at this picture")
if indicatorPos == -1 || textPos == -1 || indicatorPos >= textPos {
t.Errorf("image indicator should appear before text")
}
}
})
}
}
func TestExtractDisplayPath(t *testing.T) {
// Save original base dir
originalBaseDir := imageBaseDir
defer func() { imageBaseDir = originalBaseDir }()
tests := []struct {
name string
baseDir string
path string
expected string
}{
{
name: "no base dir shows full path",
baseDir: "",
path: "/home/user/images/cat.jpg",
expected: "/home/user/images/cat.jpg",
},
{
name: "relative path within base dir",
baseDir: "/home/user",
path: "/home/user/images/cat.jpg",
expected: "images/cat.jpg",
},
{
name: "path outside base dir shows full path",
baseDir: "/home/user",
path: "/tmp/test.jpg",
expected: "/tmp/test.jpg",
},
{
name: "same directory",
baseDir: "/home/user/images",
path: "/home/user/images/cat.jpg",
expected: "cat.jpg",
},
{
name: "long path truncated",
baseDir: "",
path: "/very/long/path/to/a/really_long_filename_that_exceeds_sixty_characters_limit_yes_it_is_very_long.jpg",
expected: "..._that_exceeds_sixty_characters_limit_yes_it_is_very_long.jpg",
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
imageBaseDir = tt.baseDir
result := extractDisplayPath(tt.path)
if result != tt.expected {
t.Errorf("extractDisplayPath(%q) with baseDir=%q = %q, want %q",
tt.path, tt.baseDir, result, tt.expected)
}
})
}
}

View File

@@ -172,3 +172,16 @@ func (orm *ORModels) ListModels(free bool) []string {
}
return resp
}
func (orm *ORModels) HasVision(modelID string) bool {
for i := range orm.Data {
if orm.Data[i].ID == modelID {
for _, mod := range orm.Data[i].Architecture.InputModalities {
if mod == "image" {
return true
}
}
}
}
return false
}

View File

@@ -109,6 +109,12 @@ func ReadCardJson(fname string) (*models.CharCard, error) {
if err := json.Unmarshal(data, &card); err != nil {
return nil, err
}
if card.FilePath == "" {
card.FilePath = fname
}
if card.ID == "" {
card.ID = models.ComputeCardID(card.Role, card.FilePath)
}
return &card, nil
}

101
popups.go
View File

@@ -1,6 +1,7 @@
package main
import (
"gf-lt/models"
"slices"
"strings"
@@ -39,9 +40,7 @@ func showModelSelectionPopup() {
default:
message = "No llama.cpp models loaded. Ensure llama.cpp server is running with models."
}
if err := notifyUser("Empty list", message); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("Empty list", message)
return
}
// Create a list primitive
@@ -51,7 +50,7 @@ func showModelSelectionPopup() {
// Find the current model index to set as selected
currentModelIndex := -1
for i, model := range modelList {
if strings.TrimPrefix(model, "(loaded) ") == chatBody.Model {
if strings.TrimPrefix(model, models.LoadedMark) == chatBody.Model {
currentModelIndex = i
}
modelListWidget.AddItem(model, "", 0, nil)
@@ -61,7 +60,7 @@ func showModelSelectionPopup() {
modelListWidget.SetCurrentItem(currentModelIndex)
}
modelListWidget.SetSelectedFunc(func(index int, mainText string, secondaryText string, shortcut rune) {
modelName := strings.TrimPrefix(mainText, "(loaded) ")
modelName := strings.TrimPrefix(mainText, models.LoadedMark)
chatBody.Model = modelName
cfg.CurrentModel = chatBody.Model
pages.RemovePage("modelSelectionPopup")
@@ -118,9 +117,7 @@ func showAPILinkSelectionPopup() {
if len(apiLinks) == 0 {
logger.Warn("no API links available for selection")
message := "No API links available. Please configure API links in your config file."
if err := notifyUser("Empty list", message); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("Empty list", message)
return
}
// Create a list primitive
@@ -142,6 +139,7 @@ func showAPILinkSelectionPopup() {
apiListWidget.SetSelectedFunc(func(index int, mainText string, secondaryText string, shortcut rune) {
// Update the API in config
cfg.CurrentAPI = mainText
// updateToolCapabilities()
// Update model list based on new API
// Helper function to get model list for a given API (same as in props_table.go)
getModelListForAPI := func(api string) []string {
@@ -159,8 +157,9 @@ func showAPILinkSelectionPopup() {
newModelList := getModelListForAPI(cfg.CurrentAPI)
// Ensure chatBody.Model is in the new list; if not, set to first available model
if len(newModelList) > 0 && !slices.Contains(newModelList, chatBody.Model) {
chatBody.Model = newModelList[0]
chatBody.Model = strings.TrimPrefix(newModelList[0], models.LoadedMark)
cfg.CurrentModel = chatBody.Model
updateToolCapabilities()
}
pages.RemovePage("apiLinkSelectionPopup")
app.SetFocus(textArea)
@@ -203,9 +202,7 @@ func showUserRoleSelectionPopup() {
if len(roles) == 0 {
logger.Warn("no roles available for selection")
message := "No roles available for selection."
if err := notifyUser("Empty list", message); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("Empty list", message)
return
}
// Create a list primitive
@@ -282,9 +279,7 @@ func showBotRoleSelectionPopup() {
if len(roles) == 0 {
logger.Warn("no roles available for selection")
message := "No roles available for selection."
if err := notifyUser("Empty list", message); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("Empty list", message)
return
}
// Create a list primitive
@@ -343,7 +338,67 @@ func showBotRoleSelectionPopup() {
app.SetFocus(roleListWidget)
}
func showFileCompletionPopup(filter string) {
func showShellFileCompletionPopup(filter string) {
baseDir := cfg.FilePickerDir
if baseDir == "" {
baseDir = "."
}
complMatches := scanFiles(baseDir, filter)
if len(complMatches) == 0 {
return
}
if len(complMatches) == 1 {
currentText := shellInput.GetText()
atIdx := strings.LastIndex(currentText, "@")
if atIdx >= 0 {
before := currentText[:atIdx]
shellInput.SetText(before + complMatches[0])
}
return
}
widget := tview.NewList().ShowSecondaryText(false).
SetSelectedBackgroundColor(tcell.ColorGray)
widget.SetTitle("file completion").SetBorder(true)
for _, m := range complMatches {
widget.AddItem(m, "", 0, nil)
}
widget.SetSelectedFunc(func(index int, mainText string, secondaryText string, shortcut rune) {
currentText := shellInput.GetText()
atIdx := strings.LastIndex(currentText, "@")
if atIdx >= 0 {
before := currentText[:atIdx]
shellInput.SetText(before + mainText)
}
pages.RemovePage("shellFileCompletionPopup")
app.SetFocus(shellInput)
})
widget.SetInputCapture(func(event *tcell.EventKey) *tcell.EventKey {
if event.Key() == tcell.KeyEscape {
pages.RemovePage("shellFileCompletionPopup")
app.SetFocus(shellInput)
return nil
}
if event.Key() == tcell.KeyRune && event.Rune() == 'x' {
pages.RemovePage("shellFileCompletionPopup")
app.SetFocus(shellInput)
return nil
}
return event
})
modal := func(p tview.Primitive, width, height int) tview.Primitive {
return tview.NewFlex().
AddItem(nil, 0, 1, false).
AddItem(tview.NewFlex().SetDirection(tview.FlexRow).
AddItem(nil, 0, 1, false).
AddItem(p, height, 1, true).
AddItem(nil, 0, 1, false), width, 1, true).
AddItem(nil, 0, 1, false)
}
pages.AddPage("shellFileCompletionPopup", modal(widget, 80, 20), true, true)
app.SetFocus(widget)
}
func showTextAreaFileCompletionPopup(filter string) {
baseDir := cfg.FilePickerDir
if baseDir == "" {
baseDir = "."
@@ -352,7 +407,6 @@ func showFileCompletionPopup(filter string) {
if len(complMatches) == 0 {
return
}
// If only one match, auto-complete without showing popup
if len(complMatches) == 1 {
currentText := textArea.GetText()
atIdx := strings.LastIndex(currentText, "@")
@@ -375,17 +429,17 @@ func showFileCompletionPopup(filter string) {
before := currentText[:atIdx]
textArea.SetText(before+mainText, true)
}
pages.RemovePage("fileCompletionPopup")
pages.RemovePage("textAreaFileCompletionPopup")
app.SetFocus(textArea)
})
widget.SetInputCapture(func(event *tcell.EventKey) *tcell.EventKey {
if event.Key() == tcell.KeyEscape {
pages.RemovePage("fileCompletionPopup")
pages.RemovePage("textAreaFileCompletionPopup")
app.SetFocus(textArea)
return nil
}
if event.Key() == tcell.KeyRune && event.Rune() == 'x' {
pages.RemovePage("fileCompletionPopup")
pages.RemovePage("textAreaFileCompletionPopup")
app.SetFocus(textArea)
return nil
}
@@ -400,8 +454,7 @@ func showFileCompletionPopup(filter string) {
AddItem(nil, 0, 1, false), width, 1, true).
AddItem(nil, 0, 1, false)
}
// Add modal page and make it visible
pages.AddPage("fileCompletionPopup", modal(widget, 80, 20), true, true)
pages.AddPage("textAreaFileCompletionPopup", modal(widget, 80, 20), true, true)
app.SetFocus(widget)
}
@@ -451,9 +504,7 @@ func showColorschemeSelectionPopup() {
if len(schemeNames) == 0 {
logger.Warn("no colorschemes available for selection")
message := "No colorschemes available."
if err := notifyUser("Empty list", message); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("Empty list", message)
return
}
// Create a list primitive

View File

@@ -259,9 +259,7 @@ func makePropsTable(props map[string]float32) *tview.Table {
// Handle nil options
if data.Options == nil {
logger.Error("options list is nil for", "label", label)
if err := notifyUser("Configuration error", "Options list is nil for "+label); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("Configuration error", "Options list is nil for "+label)
return
}
@@ -279,9 +277,7 @@ func makePropsTable(props map[string]float32) *tview.Table {
message = "No llama.cpp models loaded. Ensure llama.cpp server is running with models."
}
}
if err := notifyUser("Empty list", message); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("Empty list", message)
return
}
// Create a list primitive

View File

@@ -9,6 +9,10 @@ import (
"gf-lt/models"
"log/slog"
"net/http"
"github.com/sugarme/tokenizer"
"github.com/sugarme/tokenizer/pretrained"
"github.com/yalue/onnxruntime_go"
)
// Embedder defines the interface for embedding text
@@ -134,11 +138,150 @@ func (a *APIEmbedder) EmbedSlice(lines []string) ([][]float32, error) {
return embeddings, nil
}
// TODO: ONNXEmbedder implementation would go here
// This would require:
// 1. Loading ONNX models locally
// 2. Using a Go ONNX runtime (like gorgonia/onnx or similar)
// 3. Converting text to embeddings without external API calls
//
// For now, we'll focus on the API implementation which is already working in the current system,
// and can be extended later when we have ONNX runtime integration
type ONNXEmbedder struct {
session *onnxruntime_go.DynamicAdvancedSession
tokenizer *tokenizer.Tokenizer
dims int // embedding dimension (e.g., 768)
logger *slog.Logger
}
func NewONNXEmbedder(modelPath, tokenizerPath string, dims int, logger *slog.Logger) (*ONNXEmbedder, error) {
// Load tokenizer using sugarme/tokenizer
tok, err := pretrained.FromFile(tokenizerPath)
if err != nil {
return nil, fmt.Errorf("failed to load tokenizer: %w", err)
}
// Create ONNX session
session, err := onnxruntime_go.NewDynamicAdvancedSession(
modelPath, // onnx/embedgemma/model_q4.onnx
[]string{"input_ids", "attention_mask"},
[]string{"sentence_embedding"},
nil, // optional options
)
if err != nil {
return nil, fmt.Errorf("failed to create ONNX session: %w", err)
}
return &ONNXEmbedder{
session: session,
tokenizer: tok,
dims: dims,
logger: logger,
}, nil
}
func (e *ONNXEmbedder) Embed(text string) ([]float32, error) {
// 1. Tokenize
encoding, err := e.tokenizer.EncodeSingle(text)
if err != nil {
return nil, fmt.Errorf("tokenization failed: %w", err)
}
// 2. Convert to int64 and create attention mask
ids := encoding.Ids
inputIDs := make([]int64, len(ids))
attentionMask := make([]int64, len(ids))
for i, id := range ids {
inputIDs[i] = int64(id)
attentionMask[i] = 1
}
// 3. Create input tensors (shape: [1, seq_len])
seqLen := int64(len(inputIDs))
inputIDsTensor, err := onnxruntime_go.NewTensor[int64](
onnxruntime_go.NewShape(1, seqLen),
inputIDs,
)
if err != nil {
return nil, fmt.Errorf("failed to create input_ids tensor: %w", err)
}
defer inputIDsTensor.Destroy()
maskTensor, err := onnxruntime_go.NewTensor[int64](
onnxruntime_go.NewShape(1, seqLen),
attentionMask,
)
if err != nil {
return nil, fmt.Errorf("failed to create attention_mask tensor: %w", err)
}
defer maskTensor.Destroy()
// 4. Create output tensor
outputTensor, err := onnxruntime_go.NewEmptyTensor[float32](
onnxruntime_go.NewShape(1, int64(e.dims)),
)
if err != nil {
return nil, fmt.Errorf("failed to create output tensor: %w", err)
}
defer outputTensor.Destroy()
// 5. Run inference
err = e.session.Run(
[]onnxruntime_go.Value{inputIDsTensor, maskTensor},
[]onnxruntime_go.Value{outputTensor},
)
if err != nil {
return nil, fmt.Errorf("inference failed: %w", err)
}
// 6. Copy output data
outputData := outputTensor.GetData()
embedding := make([]float32, len(outputData))
copy(embedding, outputData)
return embedding, nil
}
func (e *ONNXEmbedder) EmbedSlice(texts []string) ([][]float32, error) {
encodings := make([]*tokenizer.Encoding, len(texts))
maxLen := 0
for i, txt := range texts {
enc, err := e.tokenizer.EncodeSingle(txt)
if err != nil {
return nil, err
}
encodings[i] = enc
if l := len(enc.Ids); l > maxLen {
maxLen = l
}
}
batchSize := len(texts)
inputIDs := make([]int64, batchSize*maxLen)
attentionMask := make([]int64, batchSize*maxLen)
for i, enc := range encodings {
ids := enc.Ids
offset := i * maxLen
for j, id := range ids {
inputIDs[offset+j] = int64(id)
attentionMask[offset+j] = 1
}
// Remaining positions are already zero (padding)
}
// Create tensors with shape [batchSize, maxLen]
inputTensor, _ := onnxruntime_go.NewTensor[int64](
onnxruntime_go.NewShape(int64(batchSize), int64(maxLen)),
inputIDs,
)
defer inputTensor.Destroy()
maskTensor, _ := onnxruntime_go.NewTensor[int64](
onnxruntime_go.NewShape(int64(batchSize), int64(maxLen)),
attentionMask,
)
defer maskTensor.Destroy()
outputTensor, _ := onnxruntime_go.NewEmptyTensor[float32](
onnxruntime_go.NewShape(int64(batchSize), int64(e.dims)),
)
defer outputTensor.Destroy()
err := e.session.Run(
[]onnxruntime_go.Value{inputTensor, maskTensor},
[]onnxruntime_go.Value{outputTensor},
)
if err != nil {
return nil, err
}
// Extract embeddings per batch item
data := outputTensor.GetData()
embeddings := make([][]float32, batchSize)
for i := 0; i < batchSize; i++ {
start := i * e.dims
emb := make([]float32, e.dims)
copy(emb, data[start:start+e.dims])
embeddings[i] = emb
}
return embeddings, nil
}

View File

@@ -246,7 +246,7 @@ func (r *RAG) extractImportantPhrases(query string) string {
break
}
}
if isImportant || len(word) > 3 {
if isImportant || len(word) >= 3 {
important = append(important, word)
}
}

104
rag_issues.md Normal file
View File

@@ -0,0 +1,104 @@
# RAG Implementation Issues and Proposed Solutions
## Overview
The current RAG system fails to retrieve relevant information for specific queries, as demonstrated by the inability to find the "two she bears" reference in the KJV Bible (2 Kings 2:23-24). While the system retrieves documents containing the word "bear", it misses the actual verse, indicating fundamental flaws in chunking, query processing, retrieval, and answer synthesis. Below we dissect each problem and propose concrete solutions.
---
### Problem 1: Chunking Destroys Semantic Coherence
- **Problem description**
The current chunking splits text into sentences and groups them by a simple word count threshold (`RAGWordLimit`). This ignores document structure (chapters, headings) and can cut through narrative units, scattering related content across multiple chunks. For the Bible query, the story of Elisha and the bears likely spans multiple verses; splitting it prevents any single chunk from containing the full context, diluting the embedding signal and making retrieval difficult.
- **Proposed solution**
- **Structure-aware chunking**: Use the EPUBs internal structure (chapters, sections) to create chunks that align with logical content units (e.g., by chapter or story).
- **Overlap between chunks**: Add a configurable overlap (e.g., 1020% of chunk size) to preserve continuity, ensuring key phrases like "two she bears" are not split across boundaries.
- **Rich metadata**: Store book name, chapter, and verse numbers with each chunk to enable filtering and source attribution.
- **Fallback to recursive splitting**: For documents without clear structure, use a recursive character text splitter with overlap (similar to LangChains `RecursiveCharacterTextSplitter`) to maintain semantic boundaries (paragraphs, sentences).
---
### Problem 2: Query Refinement Strips Critical Context
- **Problem description**
`RefineQuery` removes stop words and applies keyword-based filtering that discards semantically important modifiers. For "two she bears", the word "she" (a gender modifier) may be treated as a stop word, leaving "two bears". This loses the specificity of the query and causes the embedding to drift toward generic "bear" contexts. The rule-based approach cannot understand that "she bears" is a key phrase in the biblical story.
- **Proposed solution**
- **Entity-aware query preservation**: Use a lightweight NLP model (e.g., spaCy or a BERT-based NER tagger) to identify and retain key entities (quantities, animals, names) while only removing truly irrelevant stop words.
- **Intelligent query rewriting**: Employ a small LLM (or a set of transformation rules) to generate query variations that reflect likely biblical phrasing, e.g., "two bears came out of the wood" or "Elisha and the bears".
- **Contextual stop word removal**: Instead of a static list, use a POS tagger to keep adjectives, nouns, and verbs while removing only function words that don't carry meaning.
- **Disable refinement for short queries**: If the query is already concise (like "two she bears"), skip aggressive filtering.
---
### Problem 3: Embedding Similarity Fails for Rare or Specific Phrases
- **Problem description**
Dense embeddings excel at capturing semantic similarity but can fail when the query contains rare phrases or when the relevant passage is embedded in a noisy chunk. The verse "there came forth two she bears out of the wood" shares only the word "bears" with the query, and its embedding may be pulled toward the average of surrounding verses. Consequently, the similarity score may be lower than that of other chunks containing the word "bear" in generic contexts.
- **Proposed solution**
- **Hybrid retrieval**: Combine dense embeddings with BM25 (keyword) search. BM25 excels at exact phrase matching and would likely retrieve the verse based on "two bears" even if the embedding is weak.
- Use a library like [blevesearch](https://github.com/blevesearch/bleve) to index text alongside vectors.
- Fuse results using Reciprocal Rank Fusion (RRF) or a weighted combination.
- **Query expansion**: Add relevant terms to the query (e.g., "Elisha", "2 Kings") to improve embedding alignment.
- **Fine-tuned embeddings**: Consider using an embedding model fine-tuned on domain-specific data (e.g., biblical texts) if this is a recurring use case.
---
### Problem 4: Reranking Heuristics Are Insufficient
- **Problem description**
`RerankResults` boosts results based on simple keyword matching and file name heuristics. This coarse approach cannot reliably promote the correct verse over false positives. The adjustment `distance - score/100` is arbitrary and may not reflect true relevance.
- **Proposed solution**
- **Cross-encoder reranking**: After retrieving top candidates (e.g., top 20) with hybrid search, rerank them using a cross-encoder model that directly computes the relevance score between the query and each chunk.
- Models like `cross-encoder/ms-marco-MiniLM-L-6-v2` are lightweight and can be run locally or via a microservice.
- **Score normalization**: Use the cross-encoder scores to reorder results, discarding low-scoring ones.
- **Contextual boosting**: If metadata (e.g., chapter/verse) is available, boost results that match the querys expected location (if inferable).
---
### Problem 5: Answer Synthesis Is Not Generative
- **Problem description**
`SynthesizeAnswer` embeds a prompt and attempts to retrieve a pre-stored answer, falling back to concatenating truncated chunks. This is fundamentally flawed: RAG requires an LLM to generate a coherent answer from retrieved context. In the Bible example, even if the correct verse were retrieved, the system would only output a snippet, not an answer explaining the reference.
- **Proposed solution**
- **Integrate an LLM for generation**: Use a local model (via Ollama, Llama.cpp) or a cloud API (OpenAI, etc.) to synthesize answers.
- Construct a prompt that includes the retrieved chunks (with metadata) and the user query.
- Instruct the model to answer based solely on the provided context and cite sources (e.g., "According to 2 Kings 2:24...").
- **Implement a fallback**: If no relevant chunks are retrieved, return a message like "I couldn't find that information in your documents."
- **Streaming support**: For better UX, stream the answer token-by-token.
---
### Problem 6: Concurrency and Error Handling
- **Problem description**
The code uses a mutex only in `LoadRAG`, leaving other methods vulnerable to race conditions. The global status channel `LongJobStatusCh` may drop messages due to `select/default`, and errors are sometimes logged but not propagated. Ingestion is synchronous and slow.
- **Proposed solution**
- **Add context support**: Pass `context.Context` to all methods to allow cancellation and timeouts.
- **Worker pools for embedding**: Parallelize batch embedding with a controlled number of workers to respect API rate limits and speed up ingestion.
- **Retry logic**: Implement exponential backoff for transient API errors.
- **Replace global channel**: Use a callback or an injectable status reporter to avoid dropping messages.
- **Fine-grained locking**: Protect shared state (e.g., `storage`) with appropriate synchronization.
---
### Problem 7: Lack of Monitoring and Evaluation
- **Problem description**
There are no metrics to track retrieval quality, latency, or user satisfaction. The failure case was discovered manually; without systematic evaluation, regressions will go unnoticed.
- **Proposed solution**
- **Log key metrics**: Record query, retrieved chunk IDs, scores, and latency for each search.
- **User feedback**: Add a mechanism for users to rate answers (thumbs up/down) and use this data to improve retrieval.
- **Offline evaluation**: Create a test set of queries and expected relevant chunks (e.g., the Bible example) to measure recall@k, MRR, etc., and run it after each change.
---
## Summary
Fixing the RAG pipeline requires a multi-pronged approach:
1. **Structure-aware chunking** with metadata.
2. **Hybrid retrieval** (dense + sparse).
3. **Query understanding** via entity preservation and intelligent rewriting.
4. **Cross-encoder reranking** for precision.
5. **LLM-based answer generation**.
6. **Robust concurrency and error handling**.
7. **Monitoring and evaluation** to track improvements.
Implementing these changes will transform the system from a brittle keyword matcher into a reliable knowledge assistant capable of handling nuanced queries like the "two she bears" reference.

View File

@@ -131,13 +131,18 @@ func loadOldChatOrGetNew() []models.RoleMsg {
chat, err := store.GetLastChat()
if err != nil {
logger.Warn("failed to load history chat", "error", err)
maxID, err := store.ChatGetMaxID()
if err != nil {
logger.Error("failed to fetch max chat id", "error", err)
}
maxID++
chat := &models.Chat{
ID: 0,
ID: maxID,
CreatedAt: time.Now(),
UpdatedAt: time.Now(),
Agent: cfg.AssistantRole,
}
chat.Name = fmt.Sprintf("%s_%v", chat.Agent, chat.CreatedAt.Unix())
chat.Name = fmt.Sprintf("%s_%v", chat.Agent, chat.ID)
activeChatName = chat.Name
chatMap[chat.Name] = chat
return defaultStarter
@@ -149,10 +154,6 @@ func loadOldChatOrGetNew() []models.RoleMsg {
chatMap[chat.Name] = chat
return defaultStarter
}
// if chat.Name == "" {
// logger.Warn("empty chat name", "id", chat.ID)
// chat.Name = fmt.Sprintf("%s_%v", chat.Agent, chat.CreatedAt.Unix())
// }
chatMap[chat.Name] = chat
activeChatName = chat.Name
cfg.AssistantRole = chat.Agent
@@ -167,8 +168,3 @@ func copyToClipboard(text string) error {
cmd.Stdin = strings.NewReader(text)
return cmd.Run()
}
func notifyUser(topic, message string) error {
cmd := exec.Command("notify-send", topic, message)
return cmd.Run()
}

View File

@@ -10,16 +10,18 @@ import (
//go:embed migrations/*
var migrationsFS embed.FS
func (p *ProviderSQL) Migrate() {
func (p *ProviderSQL) Migrate() error {
// Get the embedded filesystem
migrationsDir, err := fs.Sub(migrationsFS, "migrations")
if err != nil {
p.logger.Error("Failed to get embedded migrations directory;", "error", err)
return fmt.Errorf("failed to get embedded migrations directory: %w", err)
}
// List all .up.sql files
files, err := migrationsFS.ReadDir("migrations")
if err != nil {
p.logger.Error("Failed to read migrations directory;", "error", err)
return fmt.Errorf("failed to read migrations directory: %w", err)
}
// Execute each .up.sql file
for _, file := range files {
@@ -27,11 +29,12 @@ func (p *ProviderSQL) Migrate() {
err := p.executeMigration(migrationsDir, file.Name())
if err != nil {
p.logger.Error("Failed to execute migration %s: %v", file.Name(), err)
panic(err)
return fmt.Errorf("failed to execute migration %s: %w", file.Name(), err)
}
}
}
p.logger.Debug("All migrations executed successfully!")
return nil
}
func (p *ProviderSQL) executeMigration(migrationsDir fs.FS, fileName string) error {

View File

@@ -103,7 +103,10 @@ func NewProviderSQL(dbPath string, logger *slog.Logger) FullRepo {
return nil
}
p := ProviderSQL{db: db, logger: logger}
p.Migrate()
if err := p.Migrate(); err != nil {
logger.Error("migration failed, app cannot start", "error", err)
return nil
}
return p
}

View File

@@ -1,7 +0,0 @@
{
"sys_prompt": "A game of cluedo. Players are {{user}}, {{char}}, {{char2}};\n\nrooms: hall, lounge, dinning room kitchen, ballroom, conservatory, billiard room, library, study;\nweapons: candlestick, dagger, lead pipe, revolver, rope, spanner;\npeople: miss Scarlett, colonel Mustard, mrs. White, reverend Green, mrs. Peacock, professor Plum;\n\nA murder happened in a mansion with 9 rooms. Victim is dr. Black.\nPlayers goal is to find out who commited a murder, in what room and with what weapon.\nWeapons, people and rooms not involved in murder are distributed between players (as cards) by tool agent.\nThe objective of the game is to deduce the details of the murder. There are six characters, six murder weapons, and nine rooms, leaving the players with 324 possibilities. As soon as a player enters a room, they may make a suggestion as to the details, naming a suspect, the room they are in, and the weapon. For example: \"I suspect Professor Plum, in the Dining Room, with the candlestick\".\nOnce a player makes a suggestion, the others are called upon to disprove it.\nBefore the player's move, tool agent will remind that players their cards. There are two types of moves: making a suggestion (suggestion_move) and disproving other player suggestion (evidence_move);\nIn this version player wins when the correct details are named in the suggestion_move.\n\n<example_game>\n{{user}}:\nlet's start a game of cluedo!\ntool: cards of {{char}} are 'LEAD PIPE', 'BALLROOM', 'CONSERVATORY', 'STUDY', 'Mrs. White'; suggestion_move;\n{{char}}:\n(putting miss Scarlet into the Hall with the Revolver) \"I suspect miss Scarlett, in the Hall, with the revolver.\"\ntool: cards of {{char2}} are 'SPANNER', 'DAGGER', 'Professor Plum', 'LIBRARY', 'Mrs. Peacock'; evidence_move;\n{{char2}}:\n\"No objections.\" (no cards matching the suspicion of {{char}})\ntool: cards of {{user}} are 'Colonel Mustard', 'Miss Scarlett', 'DINNING ROOM', 'CANDLESTICK', 'HALL'; evidence_move;\n{{user}}:\n\"I object. Miss Scarlett is innocent.\" (shows card with 'Miss Scarlett')\ntool: cards of {{char2}} are 'SPANNER', 'DAGGER', 'Professor Plum', 'LIBRARY', 'Mrs. Peacock'; suggestion_move;\n{{char2}}:\n*So it was not Miss Scarlett, good to know.*\n(moves Mrs. White to the Billiard Room) \"It might have been Mrs. White, in the Billiard Room, with the Revolver.\"\ntool: cards of {{user}} are 'Colonel Mustard', 'Miss Scarlett', 'DINNING ROOM', 'CANDLESTICK', 'HALL'; evidence_move;\n{{user}}:\n(no matching cards for the assumption of {{char2}}) \"Sounds possible to me.\"\ntool: cards of {{char}} are 'LEAD PIPE', 'BALLROOM', 'CONSERVATORY', 'STUDY', 'Mrs. White'; evidence_move;\n{{char}}:\n(shows Mrs. White card) \"No. Was not Mrs. White\"\ntool: cards of {{user}} are 'Colonel Mustard', 'Miss Scarlett', 'DINNING ROOM', 'CANDLESTICK', 'HALL'; suggestion_move;\n{{user}}:\n*So not Mrs. White...* (moves Reverend Green into the Billiard Room) \"I suspect Reverend Green, in the Billiard Room, with the Revolver.\"\ntool: Correct. It was Reverend Green in the Billiard Room, with the revolver. {{user}} wins.\n</example_game>",
"role": "CluedoPlayer",
"role2": "CluedoEnjoyer",
"filepath": "sysprompts/cluedo.json",
"first_msg": "Hey guys! Want to play cluedo?"
}

View File

@@ -1,6 +1,6 @@
{
"sys_prompt": "You are an expert software engineering assistant. Your goal is to help users with coding tasks, debugging, refactoring, and software development.\n\n## Core Principles\n1. **Security First**: Never expose secrets, keys, or credentials. Never commit sensitive data.\n2. **No Git Actions**: You can READ git info (status, log, diff) for context, but NEVER perform git actions (commit, add, push, checkout, reset, rm, etc.). Let the user handle all git operations.\n3. **Explore Before Execute**: Always understand the codebase structure before making changes.\n4. **Follow Conventions**: Match existing code style, patterns, and frameworks used in the project.\n5. **Be Concise**: Minimize output tokens while maintaining quality. Avoid unnecessary explanations.\n\n## Workflow for Complex Tasks\nFor multi-step tasks, ALWAYS use the todo system to track progress:\n\n1. **Create Todo List**: At the start of complex tasks, use `todo_create` to break down work into actionable items.\n2. **Update Progress**: Mark items as `in_progress` when working on them, and `completed` when done.\n3. **Check Status**: Use `todo_read` to review your progress.\n\nExample workflow:\n- User: \"Add user authentication to this app\"\n- You: Create todos: [\"Analyze existing auth structure\", \"Check frameworks in use\", \"Implement auth middleware\", \"Add login endpoints\", \"Test implementation\"]\n\n## Task Execution Flow\n\n### Phase 1: Exploration (Always First)\n- Use `file_list` to understand directory structure (path defaults to FilePickerDir if not specified)\n- Use `file_read` to examine relevant files (paths are relative to FilePickerDir unless starting with `/`)\n- Use `execute_command` with `grep`/`find` to search for patterns\n- Check `README` or documentation files\n- Identify: frameworks, conventions, testing approach\n- **Git reads allowed**: You may use `git status`, `git log`, `git diff` for context, but only to inform your work\n- **Path handling**: Relative paths are resolved against FilePickerDir (configurable via Alt+O). Use absolute paths (starting with `/`) to bypass FilePickerDir.\n\n### Phase 2: Planning\n- For complex tasks: create todo items\n- Identify files that need modification\n- Plan your approach following existing patterns\n\n### Phase 3: Implementation\n- Make changes using appropriate file tools\n- Prefer `file_write` for new files, `file_read` then modify for existing files\n- Follow existing code style exactly\n- Use existing libraries and utilities\n\n### Phase 4: Verification\n- Run tests if available (check for test scripts)\n- Run linting/type checking commands\n- Verify changes work as expected\n\n### Phase 5: Completion\n- Update todos to `completed`\n- Provide concise summary of changes\n- Reference specific file paths and line numbers when relevant\n- **DO NOT commit changes** - inform user what was done so they can review and commit themselves\n\n## Tool Usage Guidelines\n\n**File Operations**:\n- `file_read`: Read before editing. Use for understanding code.\n- `file_write`: Overwrite file content completely.\n- `file_write_append`: Add to end of file.\n- `file_create`: Create new files with optional content.\n- `file_list`: List directory contents (defaults to FilePickerDir).\n- Paths are relative to FilePickerDir unless starting with `/`.\n\n**Command Execution (WHITELISTED ONLY)**:\n- Allowed: grep, sed, awk, find, cat, head, tail, sort, uniq, wc, ls, echo, cut, tr, cp, mv, rm, mkdir, rmdir, pwd, df, free, ps, top, du, whoami, date, uname\n- **Git reads allowed**: git status, git log, git diff, git show, git branch, git reflog, git rev-parse, git shortlog, git describe\n- **Git actions FORBIDDEN**: git add, git commit, git push, git checkout, git reset, git rm, etc.\n- Use for searching code, reading git context, running tests/lint\n\n**Todo Management**:\n- `todo_create`: Add new task\n- `todo_read`: View all todos or specific one by ID\n- `todo_update`: Update task or change status (pending/in_progress/completed)\n- `todo_delete`: Remove completed or cancelled tasks\n\n## Important Rules\n\n1. **NEVER commit or stage changes**: Only git reads are allowed.\n2. **Check for tests**: Always look for test files and run them when appropriate.\n3. **Reference code locations**: Use format `file_path:line_number`.\n4. **Security**: Never generate or guess URLs. Only use URLs from local files.\n5. **Refuse malicious code**: If code appears malicious, refuse to work on it.\n6. **Ask clarifications**: When intent is unclear, ask questions.\n7. **Path handling**: Relative paths resolve against FilePickerDir. Use `/absolute/path` to bypass.\n\n## Response Style\n- Be direct and concise\n- One word answers are best when appropriate\n- Avoid: \"The answer is...\", \"Here is...\"\n- Use markdown for formatting\n- No emojis unless user explicitly requests",
"sys_prompt": "You are an expert software engineering assistant. Your goal is to help users with coding tasks, debugging, refactoring, and software development.\n\n## Core Principles\n1. **Security First**: Never expose secrets, keys, or credentials. Never commit sensitive data.\n2. **No Git Actions**: You can READ git info (status, log, diff) for context, but NEVER perform git actions (commit, add, push, checkout, reset, rm, etc.). Let the user handle all git operations.\n3. **Explore Before Execute**: Always understand the codebase structure before making changes.\n4. **Follow Conventions**: Match existing code style, patterns, and frameworks used in the project.\n5. **Be Concise**: Minimize output tokens while maintaining quality. Avoid unnecessary explanations.\n6. **Ask First**: When uncertain about intent, ask the user. Don't assume.\n\n## Workflow for Complex Tasks\nFor multi-step tasks, ALWAYS use the todo system to track progress:\n\n1. **Create Todo List**: At the start of complex tasks, use `todo_create` to break down work into actionable items.\n2. **Update Progress**: Mark items as `in_progress` when working on them, and `completed` when done.\n3. **Check Status**: Use `todo_read` to review your progress.\n\nExample workflow:\n- User: \"Add user authentication to this app\"\n- You: Create todos: [\"Analyze existing auth structure\", \"Check frameworks in use\", \"Implement auth middleware\", \"Add login endpoints\", \"Test implementation\"]\n\n## Task Execution Flow\n\n### Phase 1: Exploration (Always First)\n- Use `file_list` to understand directory structure (path defaults to FilePickerDir if not specified)\n- Use `file_read` to examine relevant files (paths are relative to FilePickerDir unless starting with `/`)\n- Use `execute_command` with `grep`/`find` to search for patterns\n- Check README, Makefile, package.json, or similar for build/test commands\n- Identify: frameworks, conventions, testing approach, lint/typecheck commands\n- **Git reads allowed**: You may use `git status`, `git log`, `git diff` for context, but only to inform your work\n- **Path handling**: Relative paths resolve against FilePickerDir; absolute paths (starting with `/`) bypass it\n\n### Phase 2: Planning\n- For complex tasks: create todo items\n- Identify files that need modification\n- Plan your approach following existing patterns\n\n### Phase 3: Implementation\n- Make changes using appropriate file tools\n- Prefer `file_write` for new files, `file_read` then edit for existing files\n- Follow existing code style exactly\n- Use existing libraries and utilities\n\n### Phase 4: Verification\n- Run tests if available (check for test scripts in README/Makefile)\n- Run linting/type checking commands\n- Verify changes work as expected\n\n### Phase 5: Completion\n- Update todos to `completed`\n- Provide concise summary of changes\n- Reference specific file paths and line numbers when relevant\n- **DO NOT commit changes** - inform user what was done so they can review and commit themselves\n\n## Command Execution\n- Use `execute_command` with a single string containing command and arguments (e.g., `go run main.go`, `ls -la`, `cd /tmp`)\n- Use `cd /path` to change the working directory for file operations",
"role": "CodingAssistant",
"filepath": "sysprompts/coding_assistant.json",
"first_msg": "Hello! I'm your coding assistant. I can help you with software engineering tasks like writing code, debugging, refactoring, and exploring codebases. I work best when you give me specific tasks, and for complex work, I'll create a todo list to track my progress. What would you like to work on?"
"first_msg": "Hello! I'm your coding assistant. Give me a specific task and I'll get started. For complex work, I'll track progress with todos."
}

111
tables.go
View File

@@ -147,9 +147,7 @@ func makeChatTable(chatMap map[string]models.Chat) *tview.Table {
if err := store.RemoveChat(sc.ID); err != nil {
logger.Error("failed to remove chat from db", "chat_id", sc.ID, "chat_name", sc.Name)
}
if err := notifyUser("chat deleted", selectedChat+" was deleted"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("chat deleted", selectedChat+" was deleted")
// load last chat
chatBody.Messages = loadOldChatOrGetNew()
textView.SetText(chatToText(chatBody.Messages, cfg.ShowSys))
@@ -159,27 +157,16 @@ func makeChatTable(chatMap map[string]models.Chat) *tview.Table {
// save updated card
fi := strings.Index(selectedChat, "_")
agentName := selectedChat[fi+1:]
cc, ok := sysMap[agentName]
if !ok {
cc := GetCardByRole(agentName)
if cc == nil {
logger.Warn("no such card", "agent", agentName)
//no:lint
if err := notifyUser("error", "no such card: "+agentName); err != nil {
logger.Warn("failed ot notify", "error", err)
}
showToast("error", "no such card: "+agentName)
return
}
// if chatBody.Messages[0].Role != "system" || chatBody.Messages[1].Role != agentName {
// if err := notifyUser("error", "unexpected chat structure; card: "+agentName); err != nil {
// logger.Warn("failed ot notify", "error", err)
// }
// return
// }
// change sys_prompt + first msg
cc.SysPrompt = chatBody.Messages[0].Content
cc.FirstMsg = chatBody.Messages[1].Content
if err := pngmeta.WriteToPng(cc.ToSpec(cfg.UserRole), cc.FilePath, cc.FilePath); err != nil {
logger.Error("failed to write charcard",
"error", err)
logger.Error("failed to write charcard", "error", err)
}
return
case "move sysprompt onto 1st msg":
@@ -190,33 +177,29 @@ func makeChatTable(chatMap map[string]models.Chat) *tview.Table {
pages.RemovePage(historyPage)
return
case "new_chat_from_card":
// Reread card from file and start fresh chat
fi := strings.Index(selectedChat, "_")
agentName := selectedChat[fi+1:]
cc, ok := sysMap[agentName]
if !ok {
cc := GetCardByRole(agentName)
if cc == nil {
logger.Warn("no such card", "agent", agentName)
if err := notifyUser("error", "no such card: "+agentName); err != nil {
logger.Warn("failed to notify", "error", err)
}
showToast("error", "no such card: "+agentName)
return
}
// Reload card from disk
newCard, err := pngmeta.ReadCard(cc.FilePath, cfg.UserRole)
if err != nil {
logger.Error("failed to reload charcard", "path", cc.FilePath, "error", err)
newCard, err = pngmeta.ReadCardJson(cc.FilePath)
if err != nil {
logger.Error("failed to reload charcard", "path", cc.FilePath, "error", err)
if err := notifyUser("error", "failed to reload card: "+cc.FilePath); err != nil {
logger.Warn("failed to notify", "error", err)
}
showToast("error", "failed to reload card: "+cc.FilePath)
return
}
}
// Update sysMap with fresh card data
sysMap[agentName] = newCard
// fetching sysprompt and first message anew from the card
if newCard.ID == "" {
newCard.ID = models.ComputeCardID(newCard.Role, newCard.FilePath)
}
sysMap[newCard.ID] = newCard
roleToID[newCard.Role] = newCard.ID
startNewChat(false)
pages.RemovePage(historyPage)
return
@@ -457,13 +440,13 @@ func makeRAGTable(fileList []string, loadedFiles []string) *tview.Flex {
go func() {
if err := ragger.LoadRAG(fpath); err != nil {
logger.Error("failed to embed file", "chat", fpath, "error", err)
_ = notifyUser("RAG", "failed to embed file; error: "+err.Error())
showToast("RAG", "failed to embed file; error: "+err.Error())
app.QueueUpdate(func() {
pages.RemovePage(RAGPage)
})
return
}
_ = notifyUser("RAG", "file loaded successfully")
showToast("RAG", "file loaded successfully")
app.QueueUpdate(func() {
pages.RemovePage(RAGPage)
})
@@ -474,13 +457,13 @@ func makeRAGTable(fileList []string, loadedFiles []string) *tview.Flex {
go func() {
if err := ragger.RemoveFile(f.name); err != nil {
logger.Error("failed to unload file from RAG", "filename", f.name, "error", err)
_ = notifyUser("RAG", "failed to unload file; error: "+err.Error())
showToast("RAG", "failed to unload file; error: "+err.Error())
app.QueueUpdate(func() {
pages.RemovePage(RAGPage)
})
return
}
_ = notifyUser("RAG", "file unloaded successfully")
showToast("RAG", "file unloaded successfully")
app.QueueUpdate(func() {
pages.RemovePage(RAGPage)
})
@@ -492,9 +475,7 @@ func makeRAGTable(fileList []string, loadedFiles []string) *tview.Flex {
logger.Error("failed to delete file", "filename", fpath, "error", err)
return
}
if err := notifyUser("chat deleted", fpath+" was deleted"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("chat deleted", fpath+" was deleted")
return
default:
pages.RemovePage(RAGPage)
@@ -529,8 +510,8 @@ func makeAgentTable(agentList []string) *tview.Table {
SetSelectable(false))
case 1:
if actions[c-1] == "filepath" {
cc, ok := sysMap[agentList[r]]
if !ok {
cc := GetCardByRole(agentList[r])
if cc == nil {
continue
}
chatActTable.SetCell(r, c,
@@ -603,9 +584,7 @@ func makeAgentTable(agentList []string) *tview.Table {
if err := store.RemoveChat(sc.ID); err != nil {
logger.Error("failed to remove chat from db", "chat_id", sc.ID, "chat_name", sc.Name)
}
if err := notifyUser("chat deleted", selected+" was deleted"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("chat deleted", selected+" was deleted")
pages.RemovePage(agentPage)
return
default:
@@ -676,13 +655,9 @@ func makeCodeBlockTable(codeBlocks []string) *tview.Table {
switch tc.Text {
case "copy":
if err := copyToClipboard(selected); err != nil {
if err := notifyUser("error", err.Error()); err != nil {
logger.Error("failed to send notification", "error", err)
}
}
if err := notifyUser("copied", selected); err != nil {
logger.Error("failed to send notification", "error", err)
showToast("error", err.Error())
}
showToast("copied", selected)
pages.RemovePage(codeBlockPage)
app.SetFocus(textArea)
return
@@ -775,9 +750,7 @@ func makeImportChatTable(filenames []string) *tview.Table {
if err := store.RemoveChat(sc.ID); err != nil {
logger.Error("failed to remove chat from db", "chat_id", sc.ID, "chat_name", sc.Name)
}
if err := notifyUser("chat deleted", selected+" was deleted"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("chat deleted", selected+" was deleted")
pages.RemovePage(historyPage)
return
default:
@@ -1130,35 +1103,11 @@ func makeFilePicker() *tview.Flex {
}
if event.Rune() == 's' {
// Set FilePickerDir to current directory
itemIndex := listView.GetCurrentItem()
if itemIndex >= 0 && itemIndex < listView.GetItemCount() {
itemText, _ := listView.GetItemText(itemIndex)
// Get the actual directory path
var targetDir string
if strings.HasPrefix(itemText, "Exit") || strings.HasPrefix(itemText, "Select this directory") {
targetDir = currentDisplayDir
} else {
actualItemName := itemText
if bracketPos := strings.Index(itemText, " ["); bracketPos != -1 {
actualItemName = itemText[:bracketPos]
}
// nolint: gocritic
if strings.HasPrefix(actualItemName, "../") {
targetDir = path.Dir(currentDisplayDir)
} else if strings.HasSuffix(actualItemName, "/") {
dirName := strings.TrimSuffix(actualItemName, "/")
targetDir = path.Join(currentDisplayDir, dirName)
} else {
targetDir = currentDisplayDir
}
}
cfg.FilePickerDir = targetDir
if err := notifyUser("FilePickerDir", "Set to: "+targetDir); err != nil {
logger.Error("failed to notify user", "error", err)
}
// pages.RemovePage(filePickerPage)
return nil
}
// Get the actual directory path
cfg.FilePickerDir = currentDisplayDir
listView.SetTitle("Files & Directories [s: set FilePickerDir]. Current base dir: " + cfg.FilePickerDir)
// pages.RemovePage(filePickerPage)
return nil
}
case tcell.KeyEnter:
// Get the currently highlighted item in the list

1028
tools.go

File diff suppressed because it is too large Load Diff

653
tools_playwright.go Normal file
View File

@@ -0,0 +1,653 @@
package main
import (
"encoding/json"
"fmt"
"gf-lt/models"
"os"
"strconv"
"strings"
"sync"
"github.com/playwright-community/playwright-go"
)
var browserToolSysMsg = `
Additional browser automation tools (Playwright):
[
{
"name": "pw_start",
"args": [],
"when_to_use": "start a browser instance before doing any browser automation. Must be called first."
},
{
"name": "pw_stop",
"args": [],
"when_to_use": "stop the browser instance when done with automation."
},
{
"name": "pw_is_running",
"args": [],
"when_to_use": "check if browser is currently running."
},
{
"name": "pw_navigate",
"args": ["url"],
"when_to_use": "open a specific URL in the web browser."
},
{
"name": "pw_click",
"args": ["selector", "index"],
"when_to_use": "click on an element on the current webpage. Use 'index' for multiple matches (default 0)."
},
{
"name": "pw_fill",
"args": ["selector", "text", "index"],
"when_to_use": "type text into an input field. Use 'index' for multiple matches (default 0)."
},
{
"name": "pw_extract_text",
"args": ["selector"],
"when_to_use": "extract text content from the page or specific elements. Use selector 'body' for all page text."
},
{
"name": "pw_screenshot",
"args": ["selector", "full_page"],
"when_to_use": "take a screenshot of the page or a specific element. Returns a file path to the image. Use to verify actions or inspect visual state."
},
{
"name": "pw_screenshot_and_view",
"args": ["selector", "full_page"],
"when_to_use": "take a screenshot and return the image for viewing. Use to visually verify page state."
},
{
"name": "pw_wait_for_selector",
"args": ["selector", "timeout"],
"when_to_use": "wait for an element to appear on the page before proceeding with further actions."
},
{
"name": "pw_drag",
"args": ["x1", "y1", "x2", "y2"],
"when_to_use": "drag the mouse from point (x1,y1) to (x2,y2)."
},
{
"name": "pw_click_at",
"args": ["x", "y"],
"when_to_use": "click at specific X,Y coordinates on the page. Use when you know the exact position."
},
{
"name": "pw_get_html",
"args": ["selector"],
"when_to_use": "get the HTML content of the page or a specific element. Use to understand page structure or extract raw HTML."
},
{
"name": "pw_get_dom",
"args": ["selector"],
"when_to_use": "get a structured DOM representation with tag, attributes, text, and children. Use to inspect element hierarchy and properties."
},
{
"name": "pw_search_elements",
"args": ["text", "selector"],
"when_to_use": "search for elements by text content or CSS selector. Returns matching elements with their tags, text, and HTML."
}
]
`
var (
pw *playwright.Playwright
browser playwright.Browser
browserStarted bool
browserStartMu sync.Mutex
page playwright.Page
)
func pwShutDown() error {
if pw == nil {
return nil
}
pwStop(nil)
return pw.Stop()
}
func installPW() error {
err := playwright.Install(&playwright.RunOptions{Verbose: false})
if err != nil {
logger.Warn("playwright not available", "error", err)
return err
}
return nil
}
func checkPlaywright() error {
var err error
pw, err = playwright.Run()
if err != nil {
logger.Warn("playwright not available", "error", err)
return err
}
return nil
}
func pwStart(args map[string]string) []byte {
browserStartMu.Lock()
defer browserStartMu.Unlock()
if browserStarted {
return []byte(`{"error": "Browser already started"}`)
}
var err error
browser, err = pw.Chromium.Launch(playwright.BrowserTypeLaunchOptions{
Headless: playwright.Bool(!cfg.PlaywrightDebug),
})
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to launch browser: %s"}`, err.Error()))
}
page, err = browser.NewPage()
if err != nil {
browser.Close()
return []byte(fmt.Sprintf(`{"error": "failed to create page: %s"}`, err.Error()))
}
browserStarted = true
return []byte(`{"success": true, "message": "Browser started"}`)
}
func pwStop(args map[string]string) []byte {
browserStartMu.Lock()
defer browserStartMu.Unlock()
if !browserStarted {
return []byte(`{"success": true, "message": "Browser was not running"}`)
}
if page != nil {
page.Close()
page = nil
}
if browser != nil {
browser.Close()
browser = nil
}
browserStarted = false
return []byte(`{"success": true, "message": "Browser stopped"}`)
}
func pwIsRunning(args map[string]string) []byte {
if browserStarted {
return []byte(`{"running": true, "message": "Browser is running"}`)
}
return []byte(`{"running": false, "message": "Browser is not running"}`)
}
func pwNavigate(args map[string]string) []byte {
url, ok := args["url"]
if !ok || url == "" {
return []byte(`{"error": "url not provided"}`)
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
_, err := page.Goto(url)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to navigate: %s"}`, err.Error()))
}
title, _ := page.Title()
pageURL := page.URL()
return []byte(fmt.Sprintf(`{"success": true, "title": "%s", "url": "%s"}`, title, pageURL))
}
func pwClick(args map[string]string) []byte {
selector, ok := args["selector"]
if !ok || selector == "" {
return []byte(`{"error": "selector not provided"}`)
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
index := 0
if args["index"] != "" {
if i, err := strconv.Atoi(args["index"]); err != nil {
logger.Warn("failed to parse index", "value", args["index"], "error", err)
} else {
index = i
}
}
locator := page.Locator(selector)
count, err := locator.Count()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to find elements: %s"}`, err.Error()))
}
if index >= count {
return []byte(fmt.Sprintf(`{"error": "Element not found at index %d (found %d elements)"}`, index, count))
}
err = locator.Nth(index).Click()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to click: %s"}`, err.Error()))
}
return []byte(`{"success": true, "message": "Clicked element"}`)
}
func pwFill(args map[string]string) []byte {
selector, ok := args["selector"]
if !ok || selector == "" {
return []byte(`{"error": "selector not provided"}`)
}
text := args["text"]
if text == "" {
text = ""
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
index := 0
if args["index"] != "" {
if i, err := strconv.Atoi(args["index"]); err != nil {
logger.Warn("failed to parse index", "value", args["index"], "error", err)
} else {
index = i
}
}
locator := page.Locator(selector)
count, err := locator.Count()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to find elements: %s"}`, err.Error()))
}
if index >= count {
return []byte(fmt.Sprintf(`{"error": "Element not found at index %d"}`, index))
}
err = locator.Nth(index).Fill(text)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to fill: %s"}`, err.Error()))
}
return []byte(`{"success": true, "message": "Filled input"}`)
}
func pwExtractText(args map[string]string) []byte {
selector := args["selector"]
if selector == "" {
selector = "body"
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
locator := page.Locator(selector)
count, err := locator.Count()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to find elements: %s"}`, err.Error()))
}
if count == 0 {
return []byte(`{"error": "No elements found"}`)
}
if selector == "body" {
text, err := page.Locator("body").TextContent()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to get text: %s"}`, err.Error()))
}
return []byte(fmt.Sprintf(`{"text": "%s"}`, text))
}
var texts []string
for i := 0; i < count; i++ {
text, err := locator.Nth(i).TextContent()
if err != nil {
continue
}
texts = append(texts, text)
}
return []byte(fmt.Sprintf(`{"text": "%s"}`, joinLines(texts)))
}
func joinLines(lines []string) string {
var sb strings.Builder
for i, line := range lines {
if i > 0 {
sb.WriteString("\n")
}
sb.WriteString(line)
}
return sb.String()
}
func pwScreenshot(args map[string]string) []byte {
selector := args["selector"]
fullPage := args["full_page"] == "true"
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
path := fmt.Sprintf("/tmp/pw_screenshot_%d.png", os.Getpid())
var err error
if selector != "" && selector != "body" {
locator := page.Locator(selector)
_, err = locator.Screenshot(playwright.LocatorScreenshotOptions{
Path: playwright.String(path),
})
} else {
_, err = page.Screenshot(playwright.PageScreenshotOptions{
Path: playwright.String(path),
FullPage: playwright.Bool(fullPage),
})
}
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to take screenshot: %s"}`, err.Error()))
}
return []byte(fmt.Sprintf(`{"path": "%s"}`, path))
}
func pwScreenshotAndView(args map[string]string) []byte {
selector := args["selector"]
fullPage := args["full_page"] == "true"
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
path := fmt.Sprintf("/tmp/pw_screenshot_%d.png", os.Getpid())
var err error
if selector != "" && selector != "body" {
locator := page.Locator(selector)
_, err = locator.Screenshot(playwright.LocatorScreenshotOptions{
Path: playwright.String(path),
})
} else {
_, err = page.Screenshot(playwright.PageScreenshotOptions{
Path: playwright.String(path),
FullPage: playwright.Bool(fullPage),
})
}
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to take screenshot: %s"}`, err.Error()))
}
dataURL, err := models.CreateImageURLFromPath(path)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to create image URL: %s"}`, err.Error()))
}
resp := models.MultimodalToolResp{
Type: "multimodal_content",
Parts: []map[string]string{
{"type": "text", "text": "Screenshot saved: " + path},
{"type": "image_url", "url": dataURL},
},
}
jsonResult, err := json.Marshal(resp)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to marshal result: %s"}`, err.Error()))
}
return jsonResult
}
func pwWaitForSelector(args map[string]string) []byte {
selector, ok := args["selector"]
if !ok || selector == "" {
return []byte(`{"error": "selector not provided"}`)
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
timeout := 30000
if args["timeout"] != "" {
if t, err := strconv.Atoi(args["timeout"]); err != nil {
logger.Warn("failed to parse timeout", "value", args["timeout"], "error", err)
} else {
timeout = t
}
}
locator := page.Locator(selector)
err := locator.WaitFor(playwright.LocatorWaitForOptions{
Timeout: playwright.Float(float64(timeout)),
})
if err != nil {
return []byte(fmt.Sprintf(`{"error": "element not found: %s"}`, err.Error()))
}
return []byte(`{"success": true, "message": "Element found"}`)
}
func pwDrag(args map[string]string) []byte {
x1, ok := args["x1"]
if !ok {
return []byte(`{"error": "x1 not provided"}`)
}
y1, ok := args["y1"]
if !ok {
return []byte(`{"error": "y1 not provided"}`)
}
x2, ok := args["x2"]
if !ok {
return []byte(`{"error": "x2 not provided"}`)
}
y2, ok := args["y2"]
if !ok {
return []byte(`{"error": "y2 not provided"}`)
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
var fx1, fy1, fx2, fy2 float64
if parsedX1, err := strconv.ParseFloat(x1, 64); err != nil {
logger.Warn("failed to parse x1", "value", x1, "error", err)
} else {
fx1 = parsedX1
}
if parsedY1, err := strconv.ParseFloat(y1, 64); err != nil {
logger.Warn("failed to parse y1", "value", y1, "error", err)
} else {
fy1 = parsedY1
}
if parsedX2, err := strconv.ParseFloat(x2, 64); err != nil {
logger.Warn("failed to parse x2", "value", x2, "error", err)
} else {
fx2 = parsedX2
}
if parsedY2, err := strconv.ParseFloat(y2, 64); err != nil {
logger.Warn("failed to parse y2", "value", y2, "error", err)
} else {
fy2 = parsedY2
}
mouse := page.Mouse()
err := mouse.Move(fx1, fy1)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to move mouse: %s"}`, err.Error()))
}
err = mouse.Down()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to mouse down: %s"}`, err.Error()))
}
err = mouse.Move(fx2, fy2)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to move mouse: %s"}`, err.Error()))
}
err = mouse.Up()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to mouse up: %s"}`, err.Error()))
}
return []byte(fmt.Sprintf(`{"success": true, "message": "Dragged from (%s,%s) to (%s,%s)"}`, x1, y1, x2, y2))
}
func pwClickAt(args map[string]string) []byte {
x, ok := args["x"]
if !ok {
return []byte(`{"error": "x not provided"}`)
}
y, ok := args["y"]
if !ok {
return []byte(`{"error": "y not provided"}`)
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
fx, err := strconv.ParseFloat(x, 64)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to parse x: %s"}`, err.Error()))
}
fy, err := strconv.ParseFloat(y, 64)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to parse y: %s"}`, err.Error()))
}
mouse := page.Mouse()
err = mouse.Click(fx, fy)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to click: %s"}`, err.Error()))
}
return []byte(fmt.Sprintf(`{"success": true, "message": "Clicked at (%s,%s)"}`, x, y))
}
func pwGetHTML(args map[string]string) []byte {
selector := args["selector"]
if selector == "" {
selector = "body"
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
locator := page.Locator(selector)
count, err := locator.Count()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to find elements: %s"}`, err.Error()))
}
if count == 0 {
return []byte(`{"error": "No elements found"}`)
}
html, err := locator.First().InnerHTML()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to get HTML: %s"}`, err.Error()))
}
return []byte(fmt.Sprintf(`{"html": %s}`, jsonString(html)))
}
type DOMElement struct {
Tag string `json:"tag,omitempty"`
Attributes map[string]string `json:"attributes,omitempty"`
Text string `json:"text,omitempty"`
Children []DOMElement `json:"children,omitempty"`
Selector string `json:"selector,omitempty"`
InnerHTML string `json:"innerHTML,omitempty"`
}
func buildDOMTree(locator playwright.Locator) ([]DOMElement, error) {
var results []DOMElement
count, err := locator.Count()
if err != nil {
return nil, err
}
for i := 0; i < count; i++ {
el := locator.Nth(i)
dom, err := elementToDOM(el)
if err != nil {
continue
}
results = append(results, dom)
}
return results, nil
}
func elementToDOM(el playwright.Locator) (DOMElement, error) {
dom := DOMElement{}
tag, err := el.Evaluate(`el => el.nodeName`, nil)
if err == nil {
dom.Tag = strings.ToLower(fmt.Sprintf("%v", tag))
}
attributes := make(map[string]string)
attrs, err := el.Evaluate(`el => {
let attrs = {};
for (let i = 0; i < el.attributes.length; i++) {
let attr = el.attributes[i];
attrs[attr.name] = attr.value;
}
return attrs;
}`, nil)
if err == nil {
if amap, ok := attrs.(map[string]any); ok {
for k, v := range amap {
if vs, ok := v.(string); ok {
attributes[k] = vs
}
}
}
}
if len(attributes) > 0 {
dom.Attributes = attributes
}
text, err := el.TextContent()
if err == nil && text != "" {
dom.Text = text
}
innerHTML, err := el.InnerHTML()
if err == nil && innerHTML != "" {
dom.InnerHTML = innerHTML
}
childCount, _ := el.Count()
if childCount > 0 {
childrenLocator := el.Locator("*")
children, err := buildDOMTree(childrenLocator)
if err == nil && len(children) > 0 {
dom.Children = children
}
}
return dom, nil
}
func pwGetDOM(args map[string]string) []byte {
selector := args["selector"]
if selector == "" {
selector = "body"
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
locator := page.Locator(selector)
count, err := locator.Count()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to find elements: %s"}`, err.Error()))
}
if count == 0 {
return []byte(`{"error": "No elements found"}`)
}
dom, err := elementToDOM(locator.First())
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to get DOM: %s"}`, err.Error()))
}
data, err := json.Marshal(dom)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to marshal DOM: %s"}`, err.Error()))
}
return []byte(fmt.Sprintf(`{"dom": %s}`, string(data)))
}
func pwSearchElements(args map[string]string) []byte {
text := args["text"]
selector := args["selector"]
if text == "" && selector == "" {
return []byte(`{"error": "text or selector not provided"}`)
}
if !browserStarted || page == nil {
return []byte(`{"error": "Browser not started. Call pw_start first."}`)
}
var locator playwright.Locator
if text != "" {
locator = page.GetByText(text)
} else {
locator = page.Locator(selector)
}
count, err := locator.Count()
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to search elements: %s"}`, err.Error()))
}
if count == 0 {
return []byte(`{"elements": []}`)
}
var results []map[string]string
for i := 0; i < count; i++ {
el := locator.Nth(i)
tag, _ := el.Evaluate(`el => el.nodeName`, nil)
text, _ := el.TextContent()
html, _ := el.InnerHTML()
results = append(results, map[string]string{
"index": strconv.Itoa(i),
"tag": strings.ToLower(fmt.Sprintf("%v", tag)),
"text": text,
"html": html,
})
}
data, err := json.Marshal(results)
if err != nil {
return []byte(fmt.Sprintf(`{"error": "failed to marshal results: %s"}`, err.Error()))
}
return []byte(fmt.Sprintf(`{"elements": %s}`, string(data)))
}
func jsonString(s string) string {
b, _ := json.Marshal(s)
return string(b)
}

452
tui.go
View File

@@ -10,6 +10,7 @@ import (
"path"
"strconv"
"strings"
"time"
"github.com/gdamore/tcell/v2"
"github.com/rivo/tview"
@@ -21,7 +22,6 @@ func isFullScreenPageActive() bool {
}
var (
app *tview.Application
pages *tview.Pages
textArea *tview.TextArea
editArea *tview.TextArea
@@ -29,11 +29,17 @@ var (
statusLineWidget *tview.TextView
helpView *tview.TextView
flex *tview.Flex
bottomFlex *tview.Flex
notificationWidget *tview.TextView
imgView *tview.Image
defaultImage = "sysprompts/llama.png"
indexPickWindow *tview.InputField
renameWindow *tview.InputField
roleEditWindow *tview.InputField
shellInput *tview.InputField
confirmModal *tview.Modal
toastTimer *time.Timer
confirmPageName = "confirm"
fullscreenMode bool
positionVisible bool = true
scrollToEndEnabled bool = true
@@ -79,7 +85,7 @@ var (
[yellow]Ctrl+p[white]: props edit form (min-p, dry, etc.)
[yellow]Ctrl+v[white]: show API link selection popup to choose current API
[yellow]Ctrl+r[white]: start/stop recording from your microphone (needs stt server or whisper binary)
[yellow]Ctrl+t[white]: remove thinking (<think>) and tool messages from context (delete from chat)
[yellow]Ctrl+t[white]: (un)collapse tool messages
[yellow]Ctrl+l[white]: show model selection popup to choose current model
[yellow]Ctrl+k[white]: switch tool use (recommend tool use to llm after user msg)
[yellow]Ctrl+a[white]: interrupt tts (needs tts server)
@@ -98,6 +104,7 @@ var (
[yellow]Alt+8[white]: show char img or last picked img
[yellow]Alt+9[white]: warm up (load) selected llama.cpp model
[yellow]Alt+t[white]: toggle thinking blocks visibility (collapse/expand <think> blocks)
[yellow]Ctrl+t[white]: toggle tool call/response visibility (collapse/expand tool calls and non-shell tool responses)
[yellow]Alt+i[white]: show colorscheme selection popup
=== scrolling chat window (some keys similar to vim) ===
@@ -124,56 +131,219 @@ Press <Enter> or 'x' to return
`
)
func setShellMode(enabled bool) {
shellMode = enabled
go func() {
app.QueueUpdateDraw(func() {
updateFlexLayout()
})
}()
}
// showToast displays a temporary notification in the bottom-right corner.
// It auto-hides after 3 seconds.
func showToast(title, message string) {
sanitize := func(s string, maxLen int) string {
sanitized := strings.Map(func(r rune) rune {
if r < 32 && r != '\t' {
return -1
}
return r
}, s)
if len(sanitized) > maxLen {
sanitized = sanitized[:maxLen-3] + "..."
}
return sanitized
}
title = sanitize(title, 50)
message = sanitize(message, 197)
if toastTimer != nil {
toastTimer.Stop()
}
// show blocking notification to not mess up flex
if fullscreenMode {
notification := tview.NewTextView().
SetTextAlign(tview.AlignCenter).
SetDynamicColors(true).
SetRegions(true).
SetText(fmt.Sprintf("[yellow]%s[-]\n", message)).
SetChangedFunc(func() {
app.Draw()
})
notification.SetTitleAlign(tview.AlignLeft).
SetBorder(true).
SetTitle(title)
// Wrap it in a fullscreen Flex to position it in the topright corner.
// Outer Flex (row) pushes content to the top; inner Flex (column) pushes to the right.
background := tview.NewFlex().SetDirection(tview.FlexRow).
AddItem(nil, 0, 1, false). // top spacer
AddItem(tview.NewFlex().SetDirection(tview.FlexColumn).
AddItem(nil, 0, 1, false). // left spacer
AddItem(notification, 40, 1, true), // notification width 40
5, 1, false) // notification height 5
// Generate a unique page name (e.g., using timestamp) to allow multiple toasts.
pageName := fmt.Sprintf("toast-%d", time.Now().UnixNano())
pages.AddPage(pageName, background, true, true)
// Autodismiss after 2 seconds, since blocking is more annoying
time.AfterFunc(2*time.Second, func() {
app.QueueUpdateDraw(func() {
if pages.HasPage(pageName) {
pages.RemovePage(pageName)
}
})
})
return
}
notificationWidget.SetTitle(title)
notificationWidget.SetText(fmt.Sprintf("[yellow]%s[-]", message))
go func() {
app.QueueUpdateDraw(func() {
flex.RemoveItem(bottomFlex)
flex.RemoveItem(statusLineWidget)
bottomFlex = tview.NewFlex().SetDirection(tview.FlexColumn).
AddItem(textArea, 0, 1, true).
AddItem(notificationWidget, 40, 1, false)
flex.AddItem(bottomFlex, 0, 10, true)
if positionVisible {
flex.AddItem(statusLineWidget, 0, 2, false)
}
})
}()
toastTimer = time.AfterFunc(3*time.Second, func() {
app.QueueUpdateDraw(func() {
flex.RemoveItem(bottomFlex)
flex.RemoveItem(statusLineWidget)
bottomFlex = tview.NewFlex().SetDirection(tview.FlexColumn).
AddItem(textArea, 0, 1, true).
AddItem(notificationWidget, 0, 0, false)
flex.AddItem(bottomFlex, 0, 10, true)
if positionVisible {
flex.AddItem(statusLineWidget, 0, 2, false)
}
})
})
}
func init() {
// Start background goroutine to update model color cache
startModelColorUpdater()
tview.Styles = colorschemes["default"]
app = tview.NewApplication()
pages = tview.NewPages()
textArea = tview.NewTextArea().
SetPlaceholder("input is multiline; press <Enter> to start the next line;\npress <Esc> to send the message.")
textArea.SetBorder(true).SetTitle("input")
// Add input capture for @ completion
textArea.SetInputCapture(func(event *tcell.EventKey) *tcell.EventKey {
shellInput = tview.NewInputField().
SetLabel(fmt.Sprintf("[%s]$ ", cfg.FilePickerDir)). // dynamic prompt
SetFieldWidth(0).
SetDoneFunc(func(key tcell.Key) {
if key == tcell.KeyEnter {
cmd := shellInput.GetText()
if cmd != "" {
executeCommandAndDisplay(cmd)
}
shellInput.SetText("")
}
})
// Copy your file completion logic to shellInput's InputCapture
shellInput.SetInputCapture(func(event *tcell.EventKey) *tcell.EventKey {
if !shellMode {
return event
}
// Handle Tab key for file completion
if event.Key() == tcell.KeyTab {
currentText := textArea.GetText()
row, col, _, _ := textArea.GetCursor()
// Calculate absolute position from row/col
lines := strings.Split(currentText, "\n")
cursorPos := 0
for i := 0; i < row && i < len(lines); i++ {
cursorPos += len(lines[i]) + 1 // +1 for newline
}
cursorPos += col
// Look backwards from cursor to find @
if cursorPos > 0 {
// Find the last @ before cursor
textBeforeCursor := currentText[:cursorPos]
atIndex := strings.LastIndex(textBeforeCursor, "@")
if atIndex >= 0 {
// Extract the partial match text after @
filter := textBeforeCursor[atIndex+1:]
showFileCompletionPopup(filter)
return nil // Consume the Tab event
// Handle Up arrow for history previous
if event.Key() == tcell.KeyUp {
if len(shellHistory) > 0 {
if shellHistoryPos < len(shellHistory)-1 {
shellHistoryPos++
shellInput.SetText(shellHistory[len(shellHistory)-1-shellHistoryPos])
}
}
return nil
}
// Handle Down arrow for history next
if event.Key() == tcell.KeyDown {
if shellHistoryPos > 0 {
shellHistoryPos--
shellInput.SetText(shellHistory[len(shellHistory)-1-shellHistoryPos])
} else if shellHistoryPos == 0 {
shellHistoryPos = -1
shellInput.SetText("")
}
return nil
}
// Reset history position when user types
if event.Key() == tcell.KeyRune {
shellHistoryPos = -1
}
// Handle Tab key for @ file completion
if event.Key() == tcell.KeyTab && shellMode {
currentText := shellInput.GetText()
atIndex := strings.LastIndex(currentText, "@")
if atIndex >= 0 {
filter := currentText[atIndex+1:]
showShellFileCompletionPopup(filter)
}
return nil
}
return event
})
confirmModal = tview.NewModal().
SetText("You are trying to send an empty message.\nIt makes sense if the last message in the chat is from you.\nAre you sure?").
AddButtons([]string{"Yes", "No"}).
SetButtonBackgroundColor(tcell.ColorBlack).
SetButtonTextColor(tcell.ColorWhite).
SetDoneFunc(func(buttonIndex int, buttonLabel string) {
if buttonLabel == "Yes" {
persona := cfg.UserRole
if cfg.WriteNextMsgAs != "" {
persona = cfg.WriteNextMsgAs
}
chatRoundChan <- &models.ChatRoundReq{Role: persona, UserMsg: ""}
} // In both Yes and No, go back to the main page
pages.SwitchToPage("main") // or whatever your main page is named
})
confirmModal.SetInputCapture(func(event *tcell.EventKey) *tcell.EventKey {
if event.Key() == tcell.KeyRune {
switch event.Rune() {
case 'y', 'Y':
persona := cfg.UserRole
if cfg.WriteNextMsgAs != "" {
persona = cfg.WriteNextMsgAs
}
chatRoundChan <- &models.ChatRoundReq{Role: persona, UserMsg: ""}
pages.SwitchToPage("main")
return nil
case 'n', 'N', 'x', 'X':
pages.SwitchToPage("main")
return nil
}
}
return event
})
textArea = tview.NewTextArea().
SetPlaceholder("input is multiline; press <Enter> to start the next line;\npress <Esc> to send the message.")
textArea.SetBorder(true).SetTitle("input")
textView = tview.NewTextView().
SetDynamicColors(true).
SetRegions(true).
SetChangedFunc(func() {
// INFO:
// https://github.com/rivo/tview/wiki/Concurrency#event-handlers
// although already called by default per tview specs
// calling it explicitly makes text streaming to look more smooth
app.Draw()
})
notificationWidget = tview.NewTextView().
SetTextAlign(tview.AlignCenter).
SetDynamicColors(true).
SetRegions(true).
SetChangedFunc(func() {
})
notificationWidget.SetBorder(true).SetTitle("notification")
bottomFlex = tview.NewFlex().SetDirection(tview.FlexColumn).
AddItem(textArea, 0, 1, true).
AddItem(notificationWidget, 0, 0, false)
//
flex = tview.NewFlex().SetDirection(tview.FlexRow).
AddItem(textView, 0, 40, false).
AddItem(textArea, 0, 10, true) // Restore original height
AddItem(bottomFlex, 0, 10, true)
if positionVisible {
flex.AddItem(statusLineWidget, 0, 2, false)
}
@@ -242,10 +412,14 @@ func init() {
// y += h / 2
// return x, y, w, h
// })
notificationWidget.SetDrawFunc(func(screen tcell.Screen, x, y, w, h int) (int, int, int, int) {
y += h / 2
return x, y, w, h
})
// Initially set up flex without search bar
flex = tview.NewFlex().SetDirection(tview.FlexRow).
AddItem(textView, 0, 40, false).
AddItem(textArea, 0, 10, true) // Restore original height
AddItem(bottomFlex, 0, 10, true)
if positionVisible {
flex.AddItem(statusLineWidget, 0, 2, false)
}
@@ -258,9 +432,7 @@ func init() {
defer colorText()
editedMsg := editArea.GetText()
if editedMsg == "" {
if err := notifyUser("edit", "no edit provided"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("edit", "no edit provided")
pages.RemovePage(editMsgPage)
return nil
}
@@ -290,9 +462,7 @@ func init() {
case tcell.KeyEnter:
newRole := roleEditWindow.GetText()
if newRole == "" {
if err := notifyUser("edit", "no role provided"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("edit", "no role provided")
pages.RemovePage(roleEditPage)
return
}
@@ -319,9 +489,7 @@ func init() {
siInt, err := strconv.Atoi(si)
if err != nil {
logger.Error("failed to convert provided index", "error", err, "si", si)
if err := notifyUser("cancel", "no index provided, copying user input"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("cancel", "no index provided, copying user input")
if err := copyToClipboard(textArea.GetText()); err != nil {
logger.Error("failed to copy to clipboard", "error", err)
}
@@ -332,9 +500,7 @@ func init() {
if len(chatBody.Messages)-1 < selectedIndex || selectedIndex < 0 {
msg := "chosen index is out of bounds, will copy user input"
logger.Warn(msg, "index", selectedIndex)
if err := notifyUser("error", msg); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("error", msg)
if err := copyToClipboard(textArea.GetText()); err != nil {
logger.Error("failed to copy to clipboard", "error", err)
}
@@ -360,9 +526,7 @@ func init() {
}
previewLen := min(30, len(msgText))
notification := fmt.Sprintf("msg '%s' was copied to the clipboard", msgText[:previewLen])
if err := notifyUser("copied", notification); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("copied", notification)
hideIndexBar() // Hide overlay after copying
}
return nil
@@ -394,9 +558,7 @@ func init() {
logger.Error("failed to upsert chat", "error", err, "chat", currentChat)
}
notification := fmt.Sprintf("renamed chat to '%s'", activeChatName)
if err := notifyUser("renamed", notification); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("renamed", notification)
}
return event
})
@@ -506,9 +668,7 @@ func init() {
if scrollToEndEnabled {
status = "enabled"
}
if err := notifyUser("autoscroll", "Auto-scrolling "+status); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("autoscroll", "Auto-scrolling "+status)
updateStatusLine()
}
// Handle Alt+7 to toggle injectRole
@@ -525,9 +685,19 @@ func init() {
if thinkingCollapsed {
status = "collapsed"
}
if err := notifyUser("thinking", "Thinking blocks "+status); err != nil {
logger.Error("failed to send notification", "error", err)
showToast("thinking", "Thinking blocks "+status)
return nil
}
// Handle Ctrl+T to toggle tool call/response visibility
if event.Key() == tcell.KeyCtrlT {
toolCollapsed = !toolCollapsed
textView.SetText(chatToText(chatBody.Messages, cfg.ShowSys))
colorText()
status := "expanded"
if toolCollapsed {
status = "collapsed"
}
showToast("tools", "Tool calls/responses "+status)
return nil
}
if event.Key() == tcell.KeyRune && event.Rune() == 'i' && event.Modifiers()&tcell.ModAlt != 0 {
@@ -547,9 +717,7 @@ func init() {
// Check if there are no chats for this agent
if len(chatList) == 0 {
notification := "no chats found for agent: " + cfg.AssistantRole
if err := notifyUser("info", notification); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("info", notification)
return nil
}
chatMap := make(map[string]models.Chat)
@@ -567,9 +735,7 @@ func init() {
if event.Key() == tcell.KeyF2 && !botRespMode {
// regen last msg
if len(chatBody.Messages) == 0 {
if err := notifyUser("info", "no messages to regenerate"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("info", "no messages to regenerate")
return nil
}
chatBody.Messages = chatBody.Messages[:len(chatBody.Messages)-1]
@@ -595,9 +761,7 @@ func init() {
return nil
}
if len(chatBody.Messages) == 0 {
if err := notifyUser("info", "no messages to delete"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("info", "no messages to delete")
return nil
}
chatBody.Messages = chatBody.Messages[:len(chatBody.Messages)-1]
@@ -643,6 +807,7 @@ func init() {
if event.Key() == tcell.KeyF6 {
interruptResp = true
botRespMode = false
toolRunningMode = false
return nil
}
if event.Key() == tcell.KeyF7 {
@@ -655,9 +820,7 @@ func init() {
}
previewLen := min(30, len(msgText))
notification := fmt.Sprintf("msg '%s' was copied to the clipboard", msgText[:previewLen])
if err := notifyUser("copied", notification); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("copied", notification)
return nil
}
if event.Key() == tcell.KeyF8 {
@@ -671,9 +834,7 @@ func init() {
text := textView.GetText(false)
cb := codeBlockRE.FindAllString(text, -1)
if len(cb) == 0 {
if err := notifyUser("notify", "no code blocks in chat"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("notify", "no code blocks in chat")
return nil
}
table := makeCodeBlockTable(cb)
@@ -688,9 +849,7 @@ func init() {
// read files in chat_exports
filelist, err := os.ReadDir(exportDir)
if err != nil {
if err := notifyUser("failed to load exports", err.Error()); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("failed to load exports", err.Error())
return nil
}
fli := []string{}
@@ -720,9 +879,7 @@ func init() {
logger.Error("failed to export chat;", "error", err, "chat_name", activeChatName)
return nil
}
if err := notifyUser("exported chat", "chat: "+activeChatName+" was exported"); err != nil {
logger.Error("failed to send notification", "error", err)
}
showToast("exported chat", "chat: "+activeChatName+" was exported")
return nil
}
if event.Key() == tcell.KeyCtrlP {
@@ -748,14 +905,6 @@ func init() {
showModelSelectionPopup()
return nil
}
if event.Key() == tcell.KeyCtrlT {
// clear context
// remove tools and thinking
removeThinking(chatBody)
textView.SetText(chatToText(chatBody.Messages, cfg.ShowSys))
colorText()
return nil
}
if event.Key() == tcell.KeyCtrlV {
if isFullScreenPageActive() {
return event
@@ -769,9 +918,7 @@ func init() {
labels, err := initSysCards()
if err != nil {
logger.Error("failed to read sys dir", "error", err)
if err := notifyUser("error", "failed to read: "+cfg.SysDir); err != nil {
logger.Debug("failed to notify user", "error", err)
}
showToast("error", "failed to read: "+cfg.SysDir)
return nil
}
at := makeAgentTable(labels)
@@ -784,6 +931,7 @@ func init() {
if event.Key() == tcell.KeyCtrlK {
// add message from tools
cfg.ToolUse = !cfg.ToolUse
updateToolCapabilities()
updateStatusLine()
return nil
}
@@ -795,21 +943,27 @@ func init() {
if err != nil {
logger.Error("failed to open attached image", "path", lastImg, "error", err)
// Fall back to showing agent image
loadImage()
if err := loadImage(); err != nil {
logger.Warn("failed to load agent image", "error", err)
}
} else {
defer file.Close()
img, _, err := image.Decode(file)
if err != nil {
logger.Error("failed to decode attached image", "path", lastImg, "error", err)
// Fall back to showing agent image
loadImage()
if err := loadImage(); err != nil {
logger.Warn("failed to load agent image", "error", err)
}
} else {
imgView.SetImage(img)
}
}
} else {
// No attached image, show agent image as before
loadImage()
if err := loadImage(); err != nil {
logger.Warn("failed to load agent image", "error", err)
}
}
pages.AddPage(imgPage, imgView, true, true)
return nil
@@ -821,9 +975,7 @@ func init() {
if err != nil {
msg := "failed to inference user speech; error:" + err.Error()
logger.Error(msg)
if err := notifyUser("stt error", msg); err != nil {
logger.Error("failed to notify user", "error", err)
}
showToast("stt error", msg)
return nil
}
if userSpeech != "" {
@@ -881,6 +1033,17 @@ func init() {
showBotRoleSelectionPopup()
return nil
}
// INFO: shutdown
if event.Key() == tcell.KeyCtrlC {
logger.Info("caught Ctrl+C via tcell event")
go func() {
if err := pwShutDown(); err != nil {
logger.Error("shutdown failed", "err", err)
}
app.Stop()
}()
return nil // swallow the event
}
if event.Key() == tcell.KeyCtrlG {
// cfg.RAGDir is the directory with files to use with RAG
// rag load
@@ -892,26 +1055,20 @@ func init() {
// Create the RAG directory if it doesn't exist
if mkdirErr := os.MkdirAll(cfg.RAGDir, 0755); mkdirErr != nil {
logger.Error("failed to create RAG directory", "dir", cfg.RAGDir, "error", mkdirErr)
if notifyerr := notifyUser("failed to create RAG directory", mkdirErr.Error()); notifyerr != nil {
logger.Error("failed to send notification", "error", notifyerr)
}
showToast("failed to create RAG directory", mkdirErr.Error())
return nil
}
// Now try to read the directory again after creating it
files, err = os.ReadDir(cfg.RAGDir)
if err != nil {
logger.Error("failed to read dir after creating it", "dir", cfg.RAGDir, "error", err)
if notifyerr := notifyUser("failed to read RAG directory", err.Error()); notifyerr != nil {
logger.Error("failed to send notification", "error", notifyerr)
}
showToast("failed to read RAG directory", err.Error())
return nil
}
} else {
// Other error (permissions, etc.)
logger.Error("failed to read dir", "dir", cfg.RAGDir, "error", err)
if notifyerr := notifyUser("failed to open RAG files dir", err.Error()); notifyerr != nil {
logger.Error("failed to send notification", "error", notifyerr)
}
showToast("failed to open RAG files dir", err.Error())
return nil
}
}
@@ -941,55 +1098,68 @@ func init() {
if event.Key() == tcell.KeyRune && event.Modifiers() == tcell.ModAlt && event.Rune() == '9' {
// Warm up (load) the currently selected model
go warmUpModel()
if err := notifyUser("model warmup", "loading model: "+chatBody.Model); err != nil {
logger.Debug("failed to notify user", "error", err)
}
showToast("model warmup", "loading model: "+chatBody.Model)
return nil
}
// cannot send msg in editMode or botRespMode
if event.Key() == tcell.KeyEscape && !editMode && !botRespMode {
msgText := textArea.GetText()
if shellMode && msgText != "" {
// In shell mode, execute command instead of sending to LLM
executeCommandAndDisplay(msgText)
textArea.SetText("", true) // Clear the input area
if shellMode {
cmdText := shellInput.GetText()
if cmdText != "" {
executeCommandAndDisplay(cmdText)
shellInput.SetText("")
}
return nil
} else if !shellMode {
// Normal mode - send to LLM
nl := "\n\n" // keep empty lines between messages
prevText := textView.GetText(true)
persona := cfg.UserRole
// strings.LastIndex()
// newline is not needed is prev msg ends with one
if strings.HasSuffix(prevText, nl) {
nl = ""
} else if strings.HasSuffix(prevText, "\n") {
nl = "\n" // only one newline, add another
}
msgText := textArea.GetText()
nl := "\n\n" // keep empty lines between messages
prevText := textView.GetText(true)
persona := cfg.UserRole
// strings.LastIndex()
// newline is not needed is prev msg ends with one
if strings.HasSuffix(prevText, nl) {
nl = ""
} else if strings.HasSuffix(prevText, "\n") {
nl = "\n" // only one newline, add another
}
if msgText != "" {
// as what char user sends msg?
if cfg.WriteNextMsgAs != "" {
persona = cfg.WriteNextMsgAs
}
if msgText != "" {
// as what char user sends msg?
if cfg.WriteNextMsgAs != "" {
persona = cfg.WriteNextMsgAs
// check if plain text
if !injectRole {
matches := roleRE.FindStringSubmatch(msgText)
if len(matches) > 1 {
persona = matches[1]
msgText = strings.TrimLeft(msgText[len(matches[0]):], " ")
}
// check if plain text
if !injectRole {
matches := roleRE.FindStringSubmatch(msgText)
if len(matches) > 1 {
persona = matches[1]
msgText = strings.TrimLeft(msgText[len(matches[0]):], " ")
}
}
// add user icon before user msg
fmt.Fprintf(textView, "%s[-:-:b](%d) <%s>: [-:-:-]\n%s\n",
nl, len(chatBody.Messages), persona, msgText)
textArea.SetText("", true)
if scrollToEndEnabled {
textView.ScrollToEnd()
}
colorText()
}
// go chatRound(msgText, persona, textView, false, false)
chatRoundChan <- &models.ChatRoundReq{Role: persona, UserMsg: msgText}
// add user icon before user msg
fmt.Fprintf(textView, "%s[-:-:b](%d) <%s>: [-:-:-]\n%s\n",
nl, len(chatBody.Messages), persona, msgText)
textArea.SetText("", true)
if scrollToEndEnabled {
textView.ScrollToEnd()
}
colorText()
} else {
pages.AddPage(confirmPageName, confirmModal, true, true)
return nil
}
// go chatRound(msgText, persona, textView, false, false)
chatRoundChan <- &models.ChatRoundReq{Role: persona, UserMsg: msgText}
return nil
}
if event.Key() == tcell.KeyTab && !shellMode {
currentF := app.GetFocus()
if currentF == textArea {
currentText := textArea.GetText()
atIndex := strings.LastIndex(currentText, "@")
if atIndex >= 0 {
filter := currentText[atIndex+1:]
showTextAreaFileCompletionPopup(filter)
}
}
return nil
}