Refactor: move msg totext method to main package
logic requires reference to config
This commit is contained in:
134
models/models.go
134
models/models.go
@@ -5,22 +5,9 @@ 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"`
|
||||
@@ -119,14 +106,14 @@ type RoleMsg struct {
|
||||
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
|
||||
//
|
||||
//nolint:gocritic
|
||||
func (m RoleMsg) MarshalJSON() ([]byte, error) {
|
||||
if m.hasContentParts {
|
||||
if m.HasContentParts {
|
||||
// Use structured content format
|
||||
aux := struct {
|
||||
Role string `json:"role"`
|
||||
@@ -189,7 +176,7 @@ func (m *RoleMsg) UnmarshalJSON(data []byte) error {
|
||||
m.IsShellCommand = structured.IsShellCommand
|
||||
m.KnownTo = structured.KnownTo
|
||||
m.Stats = structured.Stats
|
||||
m.hasContentParts = true
|
||||
m.HasContentParts = true
|
||||
return nil
|
||||
}
|
||||
|
||||
@@ -213,77 +200,13 @@ func (m *RoleMsg) UnmarshalJSON(data []byte) error {
|
||||
m.IsShellCommand = simple.IsShellCommand
|
||||
m.KnownTo = simple.KnownTo
|
||||
m.Stats = simple.Stats
|
||||
m.hasContentParts = false
|
||||
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
|
||||
@@ -316,7 +239,7 @@ func NewRoleMsg(role, content string) RoleMsg {
|
||||
return RoleMsg{
|
||||
Role: role,
|
||||
Content: content,
|
||||
hasContentParts: false,
|
||||
HasContentParts: false,
|
||||
}
|
||||
}
|
||||
|
||||
@@ -325,7 +248,7 @@ func NewMultimodalMsg(role string, contentParts []any) RoleMsg {
|
||||
return RoleMsg{
|
||||
Role: role,
|
||||
ContentParts: contentParts,
|
||||
hasContentParts: true,
|
||||
HasContentParts: true,
|
||||
}
|
||||
}
|
||||
|
||||
@@ -334,7 +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
|
||||
@@ -342,7 +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
|
||||
@@ -359,14 +282,14 @@ func (m *RoleMsg) Copy() RoleMsg {
|
||||
ToolCallID: m.ToolCallID,
|
||||
KnownTo: m.KnownTo,
|
||||
Stats: m.Stats,
|
||||
hasContentParts: m.hasContentParts,
|
||||
HasContentParts: m.HasContentParts,
|
||||
}
|
||||
}
|
||||
|
||||
// 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
|
||||
@@ -395,7 +318,7 @@ 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
|
||||
}
|
||||
@@ -425,14 +348,14 @@ 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)
|
||||
@@ -440,14 +363,14 @@ 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",
|
||||
@@ -491,31 +414,6 @@ func CreateImageURLFromPath(imagePath string) (string, error) {
|
||||
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"`
|
||||
|
||||
@@ -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)
|
||||
}
|
||||
})
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user