Feat: add rag [wip; skip-ci]

This commit is contained in:
Grail Finder
2024-12-31 13:25:13 +03:00
parent 4db8aea43d
commit 461d19aa25
13 changed files with 292 additions and 8 deletions

View File

@@ -52,3 +52,4 @@
- user icon (and role?) from config is not used; +
- message editing broke ( runtime error: index out of range [-1]); +
- F1 can load any chat, by loading chat of other agent it does not switch agents, if that chat is continued, it will rewrite agent in db; (either allow only chats from current agent OR switch agent on chat loading);
- after chat is deleted: load undeleted chat;

3
bot.go
View File

@@ -278,6 +278,9 @@ func init() {
//
logger = slog.New(slog.NewTextHandler(logfile, nil))
store = storage.NewProviderSQL("test.db", logger)
if store == nil {
os.Exit(1)
}
// https://github.com/coreydaley/ggerganov-llama.cpp/blob/master/examples/server/README.md
// load all chats in memory
if _, err := loadHistoryChats(); err != nil {

View File

@@ -8,6 +8,7 @@ import (
type Config struct {
APIURL string `toml:"APIURL"`
EmbedURL string `toml:"EmbedURL"`
ShowSys bool `toml:"ShowSys"`
LogFile string `toml:"LogFile"`
UserRole string `toml:"UserRole"`
@@ -29,6 +30,7 @@ func LoadConfigOrDefault(fn string) *Config {
if err != nil {
fmt.Println("failed to read config from file, loading default")
config.APIURL = "http://localhost:8080/v1/chat/completions"
config.EmbedURL = "http://localhost:8080/v1/embiddings"
config.ShowSys = true
config.LogFile = "log.txt"
config.UserRole = "user"

10
go.mod
View File

@@ -4,24 +4,28 @@ go 1.23.2
require (
github.com/BurntSushi/toml v1.4.0
github.com/asg017/sqlite-vec-go-bindings v0.1.6
github.com/gdamore/tcell/v2 v2.7.4
github.com/glebarez/go-sqlite v1.22.0
github.com/jmoiron/sqlx v1.4.0
github.com/ncruces/go-sqlite3 v0.21.3
github.com/rivo/tview v0.0.0-20241103174730-c76f7879f592
)
require (
github.com/dustin/go-humanize v1.0.1 // indirect
github.com/gdamore/encoding v1.0.0 // indirect
github.com/google/uuid v1.5.0 // indirect
github.com/google/uuid v1.6.0 // indirect
github.com/lucasb-eyer/go-colorful v1.2.0 // indirect
github.com/mattn/go-isatty v0.0.20 // indirect
github.com/mattn/go-runewidth v0.0.15 // indirect
github.com/ncruces/julianday v1.0.0 // indirect
github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec // indirect
github.com/rivo/uniseg v0.4.7 // indirect
golang.org/x/sys v0.17.0 // indirect
github.com/tetratelabs/wazero v1.8.2 // indirect
golang.org/x/sys v0.28.0 // indirect
golang.org/x/term v0.17.0 // indirect
golang.org/x/text v0.14.0 // indirect
golang.org/x/text v0.21.0 // indirect
modernc.org/libc v1.37.6 // indirect
modernc.org/mathutil v1.6.0 // indirect
modernc.org/memory v1.7.2 // indirect

18
go.sum
View File

@@ -2,6 +2,8 @@ filippo.io/edwards25519 v1.1.0 h1:FNf4tywRC1HmFuKW5xopWpigGjJKiJSV0Cqo0cJWDaA=
filippo.io/edwards25519 v1.1.0/go.mod h1:BxyFTGdWcka3PhytdK4V28tE5sGfRvvvRV7EaN4VDT4=
github.com/BurntSushi/toml v1.4.0 h1:kuoIxZQy2WRRk1pttg9asf+WVv6tWQuBNVmK8+nqPr0=
github.com/BurntSushi/toml v1.4.0/go.mod h1:ukJfTF/6rtPPRCnwkur4qwRxa8vTRFBF0uk2lLoLwho=
github.com/asg017/sqlite-vec-go-bindings v0.1.6 h1:Nx0jAzyS38XpkKznJ9xQjFXz2X9tI7KqjwVxV8RNoww=
github.com/asg017/sqlite-vec-go-bindings v0.1.6/go.mod h1:A8+cTt/nKFsYCQF6OgzSNpKZrzNo5gQsXBTfsXHXY0Q=
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/gdamore/encoding v1.0.0 h1:+7OoQ1Bc6eTm5niUzBa0Ctsh6JbMW6Ra+YNuAtDBdko=
@@ -14,8 +16,8 @@ github.com/go-sql-driver/mysql v1.8.1 h1:LedoTUt/eveggdHS9qUFC1EFSa8bU2+1pZjSRpv
github.com/go-sql-driver/mysql v1.8.1/go.mod h1:wEBSXgmK//2ZFJyE+qWnIsVGmvmEKlqwuVSjsCm7DZg=
github.com/google/pprof v0.0.0-20221118152302-e6195bd50e26 h1:Xim43kblpZXfIBQsbuBVKCudVG457BR2GZFIz3uw3hQ=
github.com/google/pprof v0.0.0-20221118152302-e6195bd50e26/go.mod h1:dDKJzRmX4S37WGHujM7tX//fmj1uioxKzKxz3lo4HJo=
github.com/google/uuid v1.5.0 h1:1p67kYwdtXjb0gL0BPiP1Av9wiZPo5A8z2cWkTZ+eyU=
github.com/google/uuid v1.5.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
github.com/google/uuid v1.6.0 h1:NIvaJDMOsjHA8n1jAhLSgzrAzy1Hgr+hNrb57e+94F0=
github.com/google/uuid v1.6.0/go.mod h1:TIyPZe4MgqvfeYDBFedMoGGpEw/LqOeaOT+nhxU+yHo=
github.com/jmoiron/sqlx v1.4.0 h1:1PLqN7S1UYp5t4SrVVnt4nUVNemrDAtxlulVe+Qgm3o=
github.com/jmoiron/sqlx v1.4.0/go.mod h1:ZrZ7UsYB/weZdl2Bxg6jCRO9c3YHl8r3ahlKmRT4JLY=
github.com/lib/pq v1.10.9 h1:YXG7RB+JIjhP29X+OtkiDnYaXQwpS4JEWq7dtCCRUEw=
@@ -28,6 +30,10 @@ github.com/mattn/go-runewidth v0.0.15 h1:UNAjwbU9l54TA3KzvqLGxwWjHmMgBUVhBiTjelZ
github.com/mattn/go-runewidth v0.0.15/go.mod h1:Jdepj2loyihRzMpdS35Xk/zdY8IAYHsh153qUoGf23w=
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/ncruces/go-sqlite3 v0.21.3 h1:hHkfNQLcbnxPJZhC/RGw9SwP3bfkv/Y0xUHWsr1CdMQ=
github.com/ncruces/go-sqlite3 v0.21.3/go.mod h1:zxMOaSG5kFYVFK4xQa0pdwIszqxqJ0W0BxBgwdrNjuA=
github.com/ncruces/julianday v1.0.0 h1:fH0OKwa7NWvniGQtxdJRxAgkBMolni2BjDHaWTxqt7M=
github.com/ncruces/julianday v1.0.0/go.mod h1:Dusn2KvZrrovOMJuOt0TNXL6tB7U2E8kvza5fFc9G7g=
github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec h1:W09IVJc94icq4NjY3clb7Lk8O1qJ8BdBEF8z0ibU0rE=
github.com/remyoudompheng/bigfft v0.0.0-20230129092748-24d4a6f8daec/go.mod h1:qqbHyh8v60DhA7CoWK5oRCqLrMHRGoxYCSS9EjAz6Eo=
github.com/rivo/tview v0.0.0-20241103174730-c76f7879f592 h1:YIJ+B1hePP6AgynC5TcqpO0H9k3SSoZa2BGyL6vDUzM=
@@ -36,6 +42,8 @@ github.com/rivo/uniseg v0.2.0/go.mod h1:J6wj4VEh+S6ZtnVlnTBMWIodfgj8LQOQFoIToxlJ
github.com/rivo/uniseg v0.4.3/go.mod h1:FN3SvrM+Zdj16jyLfmOkMNblXMcoc8DfTHruCPUcx88=
github.com/rivo/uniseg v0.4.7 h1:WUdvkW8uEhrYfLC4ZzdpI2ztxP1I582+49Oc5Mq64VQ=
github.com/rivo/uniseg v0.4.7/go.mod h1:FN3SvrM+Zdj16jyLfmOkMNblXMcoc8DfTHruCPUcx88=
github.com/tetratelabs/wazero v1.8.2 h1:yIgLR/b2bN31bjxwXHD8a3d+BogigR952csSDdLYEv4=
github.com/tetratelabs/wazero v1.8.2/go.mod h1:yAI0XTsMBhREkM/YDAK/zNou3GoiAce1P6+rp/wQhjs=
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=
golang.org/x/crypto v0.0.0-20210921155107-089bfa567519/go.mod h1:GvvjBRRGRdwPK5ydBHafDWAxML/pGHZbMvKqRZ5+Abc=
@@ -55,8 +63,9 @@ golang.org/x/sys v0.0.0-20220520151302-bc2c85ada10a/go.mod h1:oPkhp1MJrh7nUepCBc
golang.org/x/sys v0.0.0-20220722155257-8c9f86f7a55f/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
golang.org/x/sys v0.5.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
golang.org/x/sys v0.17.0 h1:25cE3gD+tdBA7lp7QfhuV+rJiE9YXTcS3VG1SqssI/Y=
golang.org/x/sys v0.17.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
golang.org/x/sys v0.28.0 h1:Fksou7UEQUWlKvIdsqzJmUmCX3cZuD2+P3XyyzwMhlA=
golang.org/x/sys v0.28.0/go.mod h1:/VUhepiaJMQUp4+oa/7Zr1D23ma6VTLIYjOOTFZPUcA=
golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
golang.org/x/term v0.0.0-20210927222741-03fcf44c2211/go.mod h1:jbD1KX2456YbFQfuXm/mYQcufACuNUgVhRMnK/tPxf8=
golang.org/x/term v0.5.0/go.mod h1:jMB1sMXY+tzblOD4FWmEbocvup2/aLOaQEp7JmGp78k=
@@ -66,8 +75,9 @@ golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
golang.org/x/text v0.3.7/go.mod h1:u+2+/6zg+i71rQMx5EYifcz6MCKuco9NR6JIITiCfzQ=
golang.org/x/text v0.7.0/go.mod h1:mrYo+phRRbMaCq/xk9113O4dZlRixOauAjOtrjsXDZ8=
golang.org/x/text v0.14.0 h1:ScX5w1eTa3QqT8oi6+ziP7dTV1S2+ALU0bI+0zXKWiQ=
golang.org/x/text v0.14.0/go.mod h1:18ZOQIKpY8NJVqYksKHtTdi31H5itFRjB5/qKTNYzSU=
golang.org/x/text v0.21.0 h1:zyQAAkrwaneQ066sspRyJaG9VNi/YJ1NfzcGB3hZ/qo=
golang.org/x/text v0.21.0/go.mod h1:4IBbMaMmOPCJ8SecivzSH54+73PCFmPWxNTLm+vZkEQ=
golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtnZ6UAqBI28+e2cm9otk0dWdXHAEo=
golang.org/x/tools v0.1.12/go.mod h1:hNGJHUnrk76NpqgfD5Aqm5Crs+Hm0VOH/i9J2+nxYbc=

View File

@@ -35,3 +35,13 @@ type Memory struct {
CreatedAt time.Time `db:"created_at" json:"created_at"`
UpdatedAt time.Time `db:"updated_at" json:"updated_at"`
}
// vector models
type VectorRow struct {
ID uint32 `db:"id" json:"id"`
Embeddings []float32 `db:"embeddings" json:"embeddings"`
Slug string `db:"slug" json:"slug"`
RawText string `db:"raw_text" json:"raw_text"`
Distance float32 `db:"distance" json:"distance"`
}

View File

@@ -106,3 +106,22 @@ type ChatToolsBody struct {
} `json:"tools"`
ToolChoice string `json:"tool_choice"`
}
type EmbeddingResp struct {
Embedding []float32 `json:"embedding"`
Index uint32 `json:"index"`
}
// type EmbeddingsResp struct {
// Model string `json:"model"`
// Object string `json:"object"`
// Usage struct {
// PromptTokens int `json:"prompt_tokens"`
// TotalTokens int `json:"total_tokens"`
// } `json:"usage"`
// Data []struct {
// Embedding []float32 `json:"embedding"`
// Index int `json:"index"`
// Object string `json:"object"`
// } `json:"data"`
// }

40
rag.go Normal file
View File

@@ -0,0 +1,40 @@
package main
import (
"bytes"
"elefant/models"
"encoding/json"
)
func lineToVector(line string) (*models.EmbeddingResp, error) {
payload, err := json.Marshal(map[string]string{"content": line})
if err != nil {
logger.Error("failed to marshal payload", "err:", err.Error())
return nil, err
}
resp, err := httpClient.Post(cfg.EmbedURL, "application/json", bytes.NewReader(payload))
if err != nil {
logger.Error("failed to embedd line", "err:", err.Error())
return nil, err
}
defer resp.Body.Close()
emb := models.EmbeddingResp{}
if err := json.NewDecoder(resp.Body).Decode(&emb); err != nil {
logger.Error("failed to embedd line", "err:", err.Error())
return nil, err
}
return &emb, nil
}
func saveLine(topic, line string, emb *models.EmbeddingResp) error {
row := &models.VectorRow{
Embeddings: emb.Embedding,
Slug: topic,
RawText: line,
}
return store.WriteVector(row)
}
func searchEmb(emb *models.EmbeddingResp) (*models.VectorRow, error) {
return store.SearchClosest([5120]float32(emb.Embedding))
}

View File

@@ -5,6 +5,8 @@ import (
"fmt"
"io/fs"
"strings"
_ "github.com/asg017/sqlite-vec-go-bindings/ncruces"
)
//go:embed migrations/*
@@ -27,6 +29,7 @@ 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)
}
}
}
@@ -51,7 +54,7 @@ func (p *ProviderSQL) executeMigration(migrationsDir fs.FS, fileName string) err
func (p *ProviderSQL) executeSQL(sqlContent []byte) error {
// Connect to the database (example using a simple connection)
_, err := p.db.Exec(string(sqlContent))
err := p.s3Conn.Exec(string(sqlContent))
if err != nil {
return fmt.Errorf("failed to execute SQL: %w", err)
}

View File

@@ -0,0 +1,6 @@
CREATE VIRTUAL TABLE IF NOT EXISTS embeddings USING vec0(
id INTEGER PRIMARY KEY AUTOINCREMENT,
embedding FLOAT[5120],
slug TEXT NOT NULL,
raw_text TEXT NOT NULL
);

View File

@@ -6,11 +6,13 @@ import (
_ "github.com/glebarez/go-sqlite"
"github.com/jmoiron/sqlx"
"github.com/ncruces/go-sqlite3"
)
type FullRepo interface {
ChatHistory
Memories
VectorRepo
}
type ChatHistory interface {
@@ -25,6 +27,7 @@ type ChatHistory interface {
type ProviderSQL struct {
db *sqlx.DB
s3Conn *sqlite3.Conn
logger *slog.Logger
}
@@ -87,6 +90,7 @@ func (p ProviderSQL) ChatGetMaxID() (uint32, error) {
return id, err
}
// opens two connections
func NewProviderSQL(dbPath string, logger *slog.Logger) FullRepo {
db, err := sqlx.Open("sqlite", dbPath)
if err != nil {
@@ -94,6 +98,11 @@ func NewProviderSQL(dbPath string, logger *slog.Logger) FullRepo {
return nil
}
p := ProviderSQL{db: db, logger: logger}
p.s3Conn, err = sqlite3.Open(dbPath)
if err != nil {
logger.Error("failed to open vecdb connection", "error", err)
return nil
}
p.Migrate()
return p
}

View File

@@ -3,13 +3,16 @@ package storage
import (
"elefant/models"
"fmt"
"log"
"log/slog"
"os"
"testing"
"time"
sqlite_vec "github.com/asg017/sqlite-vec-go-bindings/ncruces"
_ "github.com/glebarez/go-sqlite"
"github.com/jmoiron/sqlx"
"github.com/ncruces/go-sqlite3"
)
func TestMemories(t *testing.T) {
@@ -160,3 +163,88 @@ func TestChatHistory(t *testing.T) {
t.Errorf("Expected 0 chats, got %d", len(chats))
}
}
func TestVecTable(t *testing.T) {
// healthcheck
db, err := sqlite3.Open(":memory:")
if err != nil {
t.Fatal(err)
}
stmt, _, err := db.Prepare(`SELECT sqlite_version(), vec_version()`)
if err != nil {
t.Fatal(err)
}
stmt.Step()
log.Printf("sqlite_version=%s, vec_version=%s\n", stmt.ColumnText(0), stmt.ColumnText(1))
stmt.Close()
// migration
err = db.Exec("CREATE VIRTUAL TABLE vec_items USING vec0(embedding float[4], chat_name TEXT NOT NULL)")
if err != nil {
t.Fatal(err)
}
// data prep and insert
items := map[int][]float32{
1: {0.1, 0.1, 0.1, 0.1},
2: {0.2, 0.2, 0.2, 0.2},
3: {0.3, 0.3, 0.3, 0.3},
4: {0.4, 0.4, 0.4, 0.4},
5: {0.5, 0.5, 0.5, 0.5},
}
q := []float32{0.28, 0.3, 0.3, 0.3}
stmt, _, err = db.Prepare("INSERT INTO vec_items(rowid, embedding, chat_name) VALUES (?, ?, ?)")
if err != nil {
t.Fatal(err)
}
for id, values := range items {
v, err := sqlite_vec.SerializeFloat32(values)
if err != nil {
t.Fatal(err)
}
stmt.BindInt(1, id)
stmt.BindBlob(2, v)
stmt.BindText(3, "some_chat")
err = stmt.Exec()
if err != nil {
t.Fatal(err)
}
stmt.Reset()
}
stmt.Close()
// select | vec search
stmt, _, err = db.Prepare(`
SELECT
rowid,
distance,
embedding
FROM vec_items
WHERE embedding MATCH ?
ORDER BY distance
LIMIT 3
`)
if err != nil {
t.Fatal(err)
}
query, err := sqlite_vec.SerializeFloat32(q)
if err != nil {
t.Fatal(err)
}
stmt.BindBlob(1, query)
for stmt.Step() {
rowid := stmt.ColumnInt64(0)
distance := stmt.ColumnFloat(1)
emb := stmt.ColumnRawText(2)
floats := decodeUnsafe(emb)
log.Printf("rowid=%d, distance=%f, floats=%v\n", rowid, distance, floats)
}
if err := stmt.Err(); err != nil {
t.Fatal(err)
}
err = stmt.Close()
if err != nil {
t.Fatal(err)
}
err = db.Close()
if err != nil {
t.Fatal(err)
}
}

89
storage/vector.go Normal file
View File

@@ -0,0 +1,89 @@
package storage
import (
"elefant/models"
"fmt"
"log"
"unsafe"
sqlite_vec "github.com/asg017/sqlite-vec-go-bindings/ncruces"
)
type VectorRepo interface {
WriteVector(*models.VectorRow) error
SearchClosest(q [5120]float32) (*models.VectorRow, error)
}
var vecTableName = "embeddings"
func (p ProviderSQL) WriteVector(row *models.VectorRow) error {
stmt, _, err := p.s3Conn.Prepare(
fmt.Sprintf("INSERT INTO %s(embedding, slug, raw_text) VALUES (?, ?, ?)", vecTableName))
defer stmt.Close()
if err != nil {
p.logger.Error("failed to prep a stmt", "error", err)
return err
}
v, err := sqlite_vec.SerializeFloat32(row.Embeddings)
if err != nil {
p.logger.Error("failed to serialize vector",
"emb-len", len(row.Embeddings), "error", err)
return err
}
stmt.BindInt(1, int(row.ID))
stmt.BindBlob(2, v)
stmt.BindText(3, row.Slug)
stmt.BindText(4, row.RawText)
err = stmt.Exec()
if err != nil {
p.logger.Error("failed exec a stmt", "error", err)
return err
}
return nil
}
func decodeUnsafe(bs []byte) []float32 {
return unsafe.Slice((*float32)(unsafe.Pointer(&bs[0])), len(bs)/4)
}
func (p ProviderSQL) SearchClosest(q [5120]float32) (*models.VectorRow, error) {
stmt, _, err := p.s3Conn.Prepare(`
SELECT
id,
distance,
embedding,
slug,
raw_text
FROM vec_items
WHERE embedding MATCH ?
ORDER BY distance
LIMIT 4
`)
if err != nil {
log.Fatal(err)
}
query, err := sqlite_vec.SerializeFloat32(q[:])
if err != nil {
log.Fatal(err)
}
stmt.BindBlob(1, query)
resp := make([]models.VectorRow, 4)
i := 0
for stmt.Step() {
resp[i].ID = uint32(stmt.ColumnInt64(0))
resp[i].Distance = float32(stmt.ColumnFloat(1))
emb := stmt.ColumnRawText(2)
resp[i].Embeddings = decodeUnsafe(emb)
resp[i].Slug = stmt.ColumnText(3)
resp[i].RawText = stmt.ColumnText(4)
i++
}
if err := stmt.Err(); err != nil {
log.Fatal(err)
}
err = stmt.Close()
if err != nil {
log.Fatal(err)
}
return nil, nil
}