picoclaw/pkg/providers/openai_compat/provider.go

402 lines
11 KiB
Go
Raw Normal View History

package openai_compat
import (
"bytes"
"context"
"encoding/json"
"fmt"
"io"
"log"
"net/http"
"net/url"
"strings"
"time"
"github.com/sipeed/picoclaw/pkg/providers/protocoltypes"
)
2026-02-20 18:03:11 +00:00
type (
ToolCall = protocoltypes.ToolCall
FunctionCall = protocoltypes.FunctionCall
LLMResponse = protocoltypes.LLMResponse
UsageInfo = protocoltypes.UsageInfo
Message = protocoltypes.Message
ToolDefinition = protocoltypes.ToolDefinition
ToolFunctionDefinition = protocoltypes.ToolFunctionDefinition
ExtraContent = protocoltypes.ExtraContent
GoogleExtra = protocoltypes.GoogleExtra
ReasoningDetail = protocoltypes.ReasoningDetail
2026-02-20 18:03:11 +00:00
)
type Provider struct {
apiKey string
apiBase string
maxTokensField string // Field name for max tokens (e.g., "max_completion_tokens" for o1/glm models)
httpClient *http.Client
}
type Option func(*Provider)
const defaultRequestTimeout = 120 * time.Second
func WithMaxTokensField(maxTokensField string) Option {
return func(p *Provider) {
p.maxTokensField = maxTokensField
}
}
func WithRequestTimeout(timeout time.Duration) Option {
return func(p *Provider) {
if timeout > 0 {
p.httpClient.Timeout = timeout
}
}
}
func NewProvider(apiKey, apiBase, proxy string, opts ...Option) *Provider {
client := &http.Client{
Timeout: defaultRequestTimeout,
}
if proxy != "" {
parsed, err := url.Parse(proxy)
if err == nil {
client.Transport = &http.Transport{
Proxy: http.ProxyURL(parsed),
}
} else {
log.Printf("openai_compat: invalid proxy URL %q: %v", proxy, err)
}
}
p := &Provider{
apiKey: apiKey,
apiBase: strings.TrimRight(apiBase, "/"),
httpClient: client,
}
for _, opt := range opts {
if opt != nil {
opt(p)
}
}
return p
}
func NewProviderWithMaxTokensField(apiKey, apiBase, proxy, maxTokensField string) *Provider {
return NewProvider(apiKey, apiBase, proxy, WithMaxTokensField(maxTokensField))
}
func NewProviderWithMaxTokensFieldAndTimeout(
apiKey, apiBase, proxy, maxTokensField string,
requestTimeoutSeconds int,
) *Provider {
return NewProvider(
apiKey,
apiBase,
proxy,
WithMaxTokensField(maxTokensField),
WithRequestTimeout(time.Duration(requestTimeoutSeconds)*time.Second),
)
}
2026-02-20 18:03:11 +00:00
func (p *Provider) Chat(
ctx context.Context,
messages []Message,
tools []ToolDefinition,
model string,
options map[string]any,
) (*LLMResponse, error) {
if p.apiBase == "" {
return nil, fmt.Errorf("API base not configured")
}
model = normalizeModel(model, p.apiBase)
2026-02-20 18:03:11 +00:00
requestBody := map[string]any{
"model": model,
"messages": serializeMessages(messages),
}
if len(tools) > 0 {
requestBody["tools"] = tools
requestBody["tool_choice"] = "auto"
}
if maxTokens, ok := asInt(options["max_tokens"]); ok {
// Use configured maxTokensField if specified, otherwise fallback to model-based detection
fieldName := p.maxTokensField
if fieldName == "" {
// Fallback: detect from model name for backward compatibility
lowerModel := strings.ToLower(model)
2026-02-20 18:03:11 +00:00
if strings.Contains(lowerModel, "glm") || strings.Contains(lowerModel, "o1") ||
strings.Contains(lowerModel, "gpt-5") {
fieldName = "max_completion_tokens"
} else {
fieldName = "max_tokens"
}
}
requestBody[fieldName] = maxTokens
}
if temperature, ok := asFloat(options["temperature"]); ok {
lowerModel := strings.ToLower(model)
// Kimi k2 models only support temperature=1.
if strings.Contains(lowerModel, "kimi") && strings.Contains(lowerModel, "k2") {
requestBody["temperature"] = 1.0
} else {
requestBody["temperature"] = temperature
}
}
fix: cache system prompt with mtime-based auto-invalidation (#607) Avoid rebuilding the entire system prompt on every BuildMessages() call by caching the static portion (identity, bootstrap, skills summary, memory) and only recomputing it when workspace source files change. Key changes: - ContextBuilder caches the static prompt behind an RWMutex with double-checked locking. Source file changes are detected via cheap os.Stat mtime checks so no explicit invalidation is needed. - Track file existence at cache time (existedAtCache map) so that newly created or deleted bootstrap/memory files also trigger a rebuild — the old modifiedSince() silently returned false on os.IsNotExist. - Walk the skills directory recursively with filepath.WalkDir to catch content-only edits at any nesting depth; directory mtime alone misses in-place file modifications on most filesystems. - ToolRegistry.sortedToolNames() sorts tool names before iteration, ensuring deterministic tool definition order across calls — a prerequisite for LLM-side prefix/KV cache reuse. - Merge all context (static + dynamic + summary) into a single system message for provider compatibility: the Anthropic adapter extracts messages[0] as the top-level system parameter, and Codex reads only the first system message as instructions. - Fix a data race in BuildMessages() where cachedSystemPrompt was read without holding the lock in a debug log statement. - Add tests: single system message invariant, mtime auto-invalidation, new-file creation detection, skill file content change, explicit InvalidateCache, cache stability, concurrent access (20 goroutines x 50 iterations, passes go test -race), and a benchmark.
2026-02-25 02:34:54 +00:00
// Prompt caching: pass a stable cache key so OpenAI can bucket requests
// with the same key and reuse prefix KV cache across calls.
// The key is typically the agent ID — stable per agent, shared across requests.
// See: https://platform.openai.com/docs/guides/prompt-caching
// Prompt caching is only supported by OpenAI-native endpoints.
// Gemini and other providers reject unknown fields, so skip for non-OpenAI APIs.
fix: cache system prompt with mtime-based auto-invalidation (#607) Avoid rebuilding the entire system prompt on every BuildMessages() call by caching the static portion (identity, bootstrap, skills summary, memory) and only recomputing it when workspace source files change. Key changes: - ContextBuilder caches the static prompt behind an RWMutex with double-checked locking. Source file changes are detected via cheap os.Stat mtime checks so no explicit invalidation is needed. - Track file existence at cache time (existedAtCache map) so that newly created or deleted bootstrap/memory files also trigger a rebuild — the old modifiedSince() silently returned false on os.IsNotExist. - Walk the skills directory recursively with filepath.WalkDir to catch content-only edits at any nesting depth; directory mtime alone misses in-place file modifications on most filesystems. - ToolRegistry.sortedToolNames() sorts tool names before iteration, ensuring deterministic tool definition order across calls — a prerequisite for LLM-side prefix/KV cache reuse. - Merge all context (static + dynamic + summary) into a single system message for provider compatibility: the Anthropic adapter extracts messages[0] as the top-level system parameter, and Codex reads only the first system message as instructions. - Fix a data race in BuildMessages() where cachedSystemPrompt was read without holding the lock in a debug log statement. - Add tests: single system message invariant, mtime auto-invalidation, new-file creation detection, skill file content change, explicit InvalidateCache, cache stability, concurrent access (20 goroutines x 50 iterations, passes go test -race), and a benchmark.
2026-02-25 02:34:54 +00:00
if cacheKey, ok := options["prompt_cache_key"].(string); ok && cacheKey != "" {
if !strings.Contains(p.apiBase, "generativelanguage.googleapis.com") {
requestBody["prompt_cache_key"] = cacheKey
}
fix: cache system prompt with mtime-based auto-invalidation (#607) Avoid rebuilding the entire system prompt on every BuildMessages() call by caching the static portion (identity, bootstrap, skills summary, memory) and only recomputing it when workspace source files change. Key changes: - ContextBuilder caches the static prompt behind an RWMutex with double-checked locking. Source file changes are detected via cheap os.Stat mtime checks so no explicit invalidation is needed. - Track file existence at cache time (existedAtCache map) so that newly created or deleted bootstrap/memory files also trigger a rebuild — the old modifiedSince() silently returned false on os.IsNotExist. - Walk the skills directory recursively with filepath.WalkDir to catch content-only edits at any nesting depth; directory mtime alone misses in-place file modifications on most filesystems. - ToolRegistry.sortedToolNames() sorts tool names before iteration, ensuring deterministic tool definition order across calls — a prerequisite for LLM-side prefix/KV cache reuse. - Merge all context (static + dynamic + summary) into a single system message for provider compatibility: the Anthropic adapter extracts messages[0] as the top-level system parameter, and Codex reads only the first system message as instructions. - Fix a data race in BuildMessages() where cachedSystemPrompt was read without holding the lock in a debug log statement. - Add tests: single system message invariant, mtime auto-invalidation, new-file creation detection, skill file content change, explicit InvalidateCache, cache stability, concurrent access (20 goroutines x 50 iterations, passes go test -race), and a benchmark.
2026-02-25 02:34:54 +00:00
}
jsonData, err := json.Marshal(requestBody)
if err != nil {
return nil, fmt.Errorf("failed to marshal request: %w", err)
}
req, err := http.NewRequestWithContext(ctx, "POST", p.apiBase+"/chat/completions", bytes.NewReader(jsonData))
if err != nil {
return nil, fmt.Errorf("failed to create request: %w", err)
}
req.Header.Set("Content-Type", "application/json")
if p.apiKey != "" {
req.Header.Set("Authorization", "Bearer "+p.apiKey)
}
resp, err := p.httpClient.Do(req)
if err != nil {
return nil, fmt.Errorf("failed to send request: %w", err)
}
defer resp.Body.Close()
body, err := io.ReadAll(resp.Body)
if err != nil {
return nil, fmt.Errorf("failed to read response: %w", err)
}
if resp.StatusCode != http.StatusOK {
return nil, fmt.Errorf("API request failed:\n Status: %d\n Body: %s", resp.StatusCode, string(body))
}
return parseResponse(body)
}
func parseResponse(body []byte) (*LLMResponse, error) {
var apiResponse struct {
Choices []struct {
Message struct {
Content string `json:"content"`
ReasoningContent string `json:"reasoning_content"`
Reasoning string `json:"reasoning"`
ReasoningDetails []ReasoningDetail `json:"reasoning_details"`
ToolCalls []struct {
ID string `json:"id"`
Type string `json:"type"`
Function *struct {
Name string `json:"name"`
Arguments string `json:"arguments"`
} `json:"function"`
ExtraContent *struct {
Google *struct {
ThoughtSignature string `json:"thought_signature"`
} `json:"google"`
} `json:"extra_content"`
} `json:"tool_calls"`
} `json:"message"`
FinishReason string `json:"finish_reason"`
} `json:"choices"`
Usage *UsageInfo `json:"usage"`
}
if err := json.Unmarshal(body, &apiResponse); err != nil {
return nil, fmt.Errorf("failed to unmarshal response: %w", err)
}
if len(apiResponse.Choices) == 0 {
return &LLMResponse{
Content: "",
FinishReason: "stop",
}, nil
}
choice := apiResponse.Choices[0]
toolCalls := make([]ToolCall, 0, len(choice.Message.ToolCalls))
for _, tc := range choice.Message.ToolCalls {
2026-02-20 18:03:11 +00:00
arguments := make(map[string]any)
name := ""
// Extract thought_signature from Gemini/Google-specific extra content
thoughtSignature := ""
if tc.ExtraContent != nil && tc.ExtraContent.Google != nil {
thoughtSignature = tc.ExtraContent.Google.ThoughtSignature
}
if tc.Function != nil {
name = tc.Function.Name
if tc.Function.Arguments != "" {
if err := json.Unmarshal([]byte(tc.Function.Arguments), &arguments); err != nil {
log.Printf("openai_compat: failed to decode tool call arguments for %q: %v", name, err)
arguments["raw"] = tc.Function.Arguments
}
}
}
// Build ToolCall with ExtraContent for Gemini 3 thought_signature persistence
toolCall := ToolCall{
ID: tc.ID,
Name: name,
Arguments: arguments,
ThoughtSignature: thoughtSignature,
}
if thoughtSignature != "" {
toolCall.ExtraContent = &ExtraContent{
Google: &GoogleExtra{
ThoughtSignature: thoughtSignature,
},
}
}
toolCalls = append(toolCalls, toolCall)
}
return &LLMResponse{
Content: choice.Message.Content,
ReasoningContent: choice.Message.ReasoningContent,
Reasoning: choice.Message.Reasoning,
ReasoningDetails: choice.Message.ReasoningDetails,
ToolCalls: toolCalls,
FinishReason: choice.FinishReason,
Usage: apiResponse.Usage,
}, nil
}
fix: cache system prompt with mtime-based auto-invalidation (#607) Avoid rebuilding the entire system prompt on every BuildMessages() call by caching the static portion (identity, bootstrap, skills summary, memory) and only recomputing it when workspace source files change. Key changes: - ContextBuilder caches the static prompt behind an RWMutex with double-checked locking. Source file changes are detected via cheap os.Stat mtime checks so no explicit invalidation is needed. - Track file existence at cache time (existedAtCache map) so that newly created or deleted bootstrap/memory files also trigger a rebuild — the old modifiedSince() silently returned false on os.IsNotExist. - Walk the skills directory recursively with filepath.WalkDir to catch content-only edits at any nesting depth; directory mtime alone misses in-place file modifications on most filesystems. - ToolRegistry.sortedToolNames() sorts tool names before iteration, ensuring deterministic tool definition order across calls — a prerequisite for LLM-side prefix/KV cache reuse. - Merge all context (static + dynamic + summary) into a single system message for provider compatibility: the Anthropic adapter extracts messages[0] as the top-level system parameter, and Codex reads only the first system message as instructions. - Fix a data race in BuildMessages() where cachedSystemPrompt was read without holding the lock in a debug log statement. - Add tests: single system message invariant, mtime auto-invalidation, new-file creation detection, skill file content change, explicit InvalidateCache, cache stability, concurrent access (20 goroutines x 50 iterations, passes go test -race), and a benchmark.
2026-02-25 02:34:54 +00:00
// openaiMessage is the wire-format message for OpenAI-compatible APIs.
// It mirrors protocoltypes.Message but omits SystemParts, which is an
// internal field that would be unknown to third-party endpoints.
type openaiMessage struct {
Role string `json:"role"`
Content string `json:"content"`
ReasoningContent string `json:"reasoning_content,omitempty"`
ToolCalls []ToolCall `json:"tool_calls,omitempty"`
ToolCallID string `json:"tool_call_id,omitempty"`
fix: cache system prompt with mtime-based auto-invalidation (#607) Avoid rebuilding the entire system prompt on every BuildMessages() call by caching the static portion (identity, bootstrap, skills summary, memory) and only recomputing it when workspace source files change. Key changes: - ContextBuilder caches the static prompt behind an RWMutex with double-checked locking. Source file changes are detected via cheap os.Stat mtime checks so no explicit invalidation is needed. - Track file existence at cache time (existedAtCache map) so that newly created or deleted bootstrap/memory files also trigger a rebuild — the old modifiedSince() silently returned false on os.IsNotExist. - Walk the skills directory recursively with filepath.WalkDir to catch content-only edits at any nesting depth; directory mtime alone misses in-place file modifications on most filesystems. - ToolRegistry.sortedToolNames() sorts tool names before iteration, ensuring deterministic tool definition order across calls — a prerequisite for LLM-side prefix/KV cache reuse. - Merge all context (static + dynamic + summary) into a single system message for provider compatibility: the Anthropic adapter extracts messages[0] as the top-level system parameter, and Codex reads only the first system message as instructions. - Fix a data race in BuildMessages() where cachedSystemPrompt was read without holding the lock in a debug log statement. - Add tests: single system message invariant, mtime auto-invalidation, new-file creation detection, skill file content change, explicit InvalidateCache, cache stability, concurrent access (20 goroutines x 50 iterations, passes go test -race), and a benchmark.
2026-02-25 02:34:54 +00:00
}
// serializeMessages converts internal Message structs to the OpenAI wire format.
// - Strips SystemParts (unknown to third-party endpoints)
// - Converts messages with Media to multipart content format (text + image_url parts)
// - Preserves ToolCallID, ToolCalls, and ReasoningContent for all messages
func serializeMessages(messages []Message) []any {
out := make([]any, 0, len(messages))
for _, m := range messages {
if len(m.Media) == 0 {
out = append(out, openaiMessage{
Role: m.Role,
Content: m.Content,
ReasoningContent: m.ReasoningContent,
ToolCalls: m.ToolCalls,
ToolCallID: m.ToolCallID,
})
continue
fix: cache system prompt with mtime-based auto-invalidation (#607) Avoid rebuilding the entire system prompt on every BuildMessages() call by caching the static portion (identity, bootstrap, skills summary, memory) and only recomputing it when workspace source files change. Key changes: - ContextBuilder caches the static prompt behind an RWMutex with double-checked locking. Source file changes are detected via cheap os.Stat mtime checks so no explicit invalidation is needed. - Track file existence at cache time (existedAtCache map) so that newly created or deleted bootstrap/memory files also trigger a rebuild — the old modifiedSince() silently returned false on os.IsNotExist. - Walk the skills directory recursively with filepath.WalkDir to catch content-only edits at any nesting depth; directory mtime alone misses in-place file modifications on most filesystems. - ToolRegistry.sortedToolNames() sorts tool names before iteration, ensuring deterministic tool definition order across calls — a prerequisite for LLM-side prefix/KV cache reuse. - Merge all context (static + dynamic + summary) into a single system message for provider compatibility: the Anthropic adapter extracts messages[0] as the top-level system parameter, and Codex reads only the first system message as instructions. - Fix a data race in BuildMessages() where cachedSystemPrompt was read without holding the lock in a debug log statement. - Add tests: single system message invariant, mtime auto-invalidation, new-file creation detection, skill file content change, explicit InvalidateCache, cache stability, concurrent access (20 goroutines x 50 iterations, passes go test -race), and a benchmark.
2026-02-25 02:34:54 +00:00
}
// Multipart content format for messages with media
parts := make([]map[string]any, 0, 1+len(m.Media))
if m.Content != "" {
parts = append(parts, map[string]any{
"type": "text",
"text": m.Content,
})
}
for _, mediaURL := range m.Media {
2026-03-04 18:34:08 +00:00
if strings.HasPrefix(mediaURL, "data:image/") {
parts = append(parts, map[string]any{
"type": "image_url",
"image_url": map[string]any{
"url": mediaURL,
},
})
}
}
msg := map[string]any{
"role": m.Role,
"content": parts,
}
if m.ToolCallID != "" {
msg["tool_call_id"] = m.ToolCallID
}
if len(m.ToolCalls) > 0 {
msg["tool_calls"] = m.ToolCalls
}
if m.ReasoningContent != "" {
msg["reasoning_content"] = m.ReasoningContent
}
out = append(out, msg)
fix: cache system prompt with mtime-based auto-invalidation (#607) Avoid rebuilding the entire system prompt on every BuildMessages() call by caching the static portion (identity, bootstrap, skills summary, memory) and only recomputing it when workspace source files change. Key changes: - ContextBuilder caches the static prompt behind an RWMutex with double-checked locking. Source file changes are detected via cheap os.Stat mtime checks so no explicit invalidation is needed. - Track file existence at cache time (existedAtCache map) so that newly created or deleted bootstrap/memory files also trigger a rebuild — the old modifiedSince() silently returned false on os.IsNotExist. - Walk the skills directory recursively with filepath.WalkDir to catch content-only edits at any nesting depth; directory mtime alone misses in-place file modifications on most filesystems. - ToolRegistry.sortedToolNames() sorts tool names before iteration, ensuring deterministic tool definition order across calls — a prerequisite for LLM-side prefix/KV cache reuse. - Merge all context (static + dynamic + summary) into a single system message for provider compatibility: the Anthropic adapter extracts messages[0] as the top-level system parameter, and Codex reads only the first system message as instructions. - Fix a data race in BuildMessages() where cachedSystemPrompt was read without holding the lock in a debug log statement. - Add tests: single system message invariant, mtime auto-invalidation, new-file creation detection, skill file content change, explicit InvalidateCache, cache stability, concurrent access (20 goroutines x 50 iterations, passes go test -race), and a benchmark.
2026-02-25 02:34:54 +00:00
}
return out
}
func normalizeModel(model, apiBase string) string {
before, after, ok := strings.Cut(model, "/")
if !ok {
return model
}
if strings.Contains(strings.ToLower(apiBase), "openrouter.ai") {
return model
}
prefix := strings.ToLower(before)
switch prefix {
case "litellm", "moonshot", "nvidia", "groq", "ollama", "deepseek", "google", "openrouter", "zhipu", "mistral":
return after
default:
return model
}
}
2026-02-20 18:03:11 +00:00
func asInt(v any) (int, bool) {
switch val := v.(type) {
case int:
return val, true
case int64:
return int(val), true
case float64:
return int(val), true
case float32:
return int(val), true
default:
return 0, false
}
}
2026-02-20 18:03:11 +00:00
func asFloat(v any) (float64, bool) {
switch val := v.(type) {
case float64:
return val, true
case float32:
return float64(val), true
case int:
return float64(val), true
case int64:
return float64(val), true
default:
return 0, false
}
}