MCP Memory LibSQL Go

package embeddings import ( "bytes" "context" "encoding/json" "fmt" "net/http" "os" "strings" "time" ) // Vertex AI Text Embeddings API (REST) via AIP // Endpoint pattern: // https://{location}-aiplatform.googleapis.com/v1/projects/{project}/locations/{location}/publishers/google/models/{model}:predict // FIXME: Requires OAuth2. For simplicity, we allow a direct endpoint+token via env in this implementation. type vertexProvider struct { endpoint string // full URL to :predict token string // OAuth2 access token (bearer) dims int http *http.Client } func newVertexFromEnv() Provider { endpoint := strings.TrimSpace(os.Getenv("VERTEX_EMBEDDINGS_ENDPOINT")) token := strings.TrimSpace(os.Getenv("VERTEX_ACCESS_TOKEN")) if endpoint == "" || token == "" { return nil } dims := 768 return &vertexProvider{endpoint: endpoint, token: token, dims: dims, http: &http.Client{Timeout: 15 * time.Second}} } func (p *vertexProvider) Name() string { return "vertexai" } func (p *vertexProvider) Dimensions() int { return p.dims } func (p *vertexProvider) Embed(ctx context.Context, inputs []string) ([][]float32, error) { if len(inputs) == 0 { return [][]float32{}, nil } // Vertex predict supports batch payloads depending on model; to keep simple, do per-input. res := make([][]float32, 0, len(inputs)) for _, in := range inputs { payload := map[string]any{ "instances": []any{map[string]any{"content": in}}, } b, _ := json.Marshal(payload) req, err := http.NewRequestWithContext(ctx, http.MethodPost, p.endpoint, bytes.NewReader(b)) if err != nil { return nil, err } req.Header.Set("Content-Type", "application/json") req.Header.Set("Authorization", "Bearer "+p.token) resp, err := p.http.Do(req) if err != nil { return nil, err } var out struct { Predictions []struct { Embeddings struct { Values []float64 `json:"values"` } `json:"embeddings"` } `json:"predictions"` } if resp.StatusCode < 200 || resp.StatusCode >= 300 { var er struct { Error struct { Message string `json:"message"` } `json:"error"` } _ = json.NewDecoder(resp.Body).Decode(&er) resp.Body.Close() if er.Error.Message != "" { return nil, fmt.Errorf("vertex error: %s", er.Error.Message) } return nil, fmt.Errorf("vertex http status: %s", resp.Status) } if err := json.NewDecoder(resp.Body).Decode(&out); err != nil { resp.Body.Close() return nil, err } resp.Body.Close() if len(out.Predictions) == 0 { return nil, fmt.Errorf("vertex returned no predictions") } res = append(res, f64to32(out.Predictions[0].Embeddings.Values)) } return res, nil }