picoclaw/pkg/providers/anthropic/provider.go
Amir Mamaghani 468723be85 feat(telegram): stream LLM responses in real-time via sendMessageDraft
Implements real-time token streaming to Telegram using the sendMessageDraft
API (telego v1.6.0). Instead of showing only a "Thinking..." placeholder
until the full response arrives, users now see partial LLM output appear
in the chat as it's generated.

The streaming pipeline threads through all layers:

- StreamingProvider interface (providers/types.go): opt-in ChatStream()
  method that receives an onChunk callback with accumulated text
- OpenAI-compatible SSE streaming (openai_compat/provider.go): parses
  SSE events with stream:true, handles text deltas and tool call assembly
- Anthropic native streaming (anthropic/provider.go): uses SDK's
  NewStreaming() for direct Anthropic API connections
- HTTPProvider delegation (http_provider.go): delegates ChatStream to
  the underlying openai_compat provider
- StreamingCapable + Streamer interfaces (channels/interfaces.go):
  opt-in channel capability like TypingCapable/PlaceholderCapable
- Telegram streamer (telegram/telegram.go): BeginStream returns a
  telegramStreamer that throttles sendMessageDraft calls (3s/200 chars)
  with graceful degradation on API errors
- StreamDelegate bridge (bus/bus.go): decouples agent loop from channel
  manager without tight imports
- Manager integration (manager.go): implements StreamDelegate, tracks
  streamActive state, coordinates with placeholder editing
- Agent loop (loop.go): uses ChatStream when both provider and channel
  support streaming, cancels stream on tool calls, skips PublishOutbound
  when Finalize already delivered the message

Graceful degradation:
- Bots without forum/topics mode: first sendMessageDraft error sets
  failed=true, subsequent Updates become no-ops, Finalize still delivers
  via SendMessage. User sees normal non-streaming behavior.
- Non-streaming providers: type assertion fails, falls back to Chat()
- Config opt-out: streaming.enabled (default true) in telegram config

Closes #1098
2026-03-04 19:19:25 +01:00

406 lines
10 KiB
Go

package anthropicprovider
import (
"context"
"encoding/json"
"fmt"
"log"
"strings"
"github.com/anthropics/anthropic-sdk-go"
"github.com/anthropics/anthropic-sdk-go/option"
"github.com/anthropics/anthropic-sdk-go/packages/ssestream"
"github.com/sipeed/picoclaw/pkg/providers/protocoltypes"
)
type (
ToolCall = protocoltypes.ToolCall
FunctionCall = protocoltypes.FunctionCall
LLMResponse = protocoltypes.LLMResponse
UsageInfo = protocoltypes.UsageInfo
Message = protocoltypes.Message
ToolDefinition = protocoltypes.ToolDefinition
ToolFunctionDefinition = protocoltypes.ToolFunctionDefinition
)
const defaultBaseURL = "https://api.anthropic.com"
type Provider struct {
client *anthropic.Client
tokenSource func() (string, error)
baseURL string
}
func NewProvider(token string) *Provider {
return NewProviderWithBaseURL(token, "")
}
func NewProviderWithBaseURL(token, apiBase string) *Provider {
baseURL := normalizeBaseURL(apiBase)
client := anthropic.NewClient(
option.WithAuthToken(token),
option.WithBaseURL(baseURL),
)
return &Provider{
client: &client,
baseURL: baseURL,
}
}
func NewProviderWithClient(client *anthropic.Client) *Provider {
return &Provider{
client: client,
baseURL: defaultBaseURL,
}
}
func NewProviderWithTokenSource(token string, tokenSource func() (string, error)) *Provider {
return NewProviderWithTokenSourceAndBaseURL(token, tokenSource, "")
}
func NewProviderWithTokenSourceAndBaseURL(token string, tokenSource func() (string, error), apiBase string) *Provider {
p := NewProviderWithBaseURL(token, apiBase)
p.tokenSource = tokenSource
return p
}
func (p *Provider) Chat(
ctx context.Context,
messages []Message,
tools []ToolDefinition,
model string,
options map[string]any,
) (*LLMResponse, error) {
var opts []option.RequestOption
if p.tokenSource != nil {
tok, err := p.tokenSource()
if err != nil {
return nil, fmt.Errorf("refreshing token: %w", err)
}
opts = append(opts, option.WithAuthToken(tok))
}
params, err := buildParams(messages, tools, model, options)
if err != nil {
return nil, err
}
resp, err := p.client.Messages.New(ctx, params, opts...)
if err != nil {
return nil, fmt.Errorf("claude API call: %w", err)
}
return parseResponse(resp), nil
}
// ChatStream implements providers.StreamingProvider.
// It streams text deltas to onChunk (with accumulated text) while building
// the same LLMResponse returned by Chat for tool-call compatibility.
func (p *Provider) ChatStream(
ctx context.Context,
messages []Message,
tools []ToolDefinition,
model string,
options map[string]any,
onChunk func(accumulated string),
) (*LLMResponse, error) {
var opts []option.RequestOption
if p.tokenSource != nil {
tok, err := p.tokenSource()
if err != nil {
return nil, fmt.Errorf("refreshing token: %w", err)
}
opts = append(opts, option.WithAuthToken(tok))
}
params, err := buildParams(messages, tools, model, options)
if err != nil {
return nil, err
}
stream := p.client.Messages.NewStreaming(ctx, params, opts...)
return parseStream(stream, onChunk)
}
// parseStream consumes a streaming response and builds an LLMResponse.
func parseStream(
stream *ssestream.Stream[anthropic.MessageStreamEventUnion],
onChunk func(accumulated string),
) (*LLMResponse, error) {
defer stream.Close()
var textContent strings.Builder
var toolCalls []ToolCall
var stopReason anthropic.StopReason
var inputTokens, outputTokens int64
// Track tool_use blocks being assembled (by content block index)
type toolBlock struct {
id string
name string
inputJSON strings.Builder
}
activeTools := map[int64]*toolBlock{}
for stream.Next() {
event := stream.Current()
switch event.Type {
case "message_start":
if event.Message.Usage.InputTokens > 0 {
inputTokens = event.Message.Usage.InputTokens
}
case "content_block_start":
cb := event.ContentBlock
if cb.Type == "tool_use" {
activeTools[event.Index] = &toolBlock{
id: cb.ID,
name: cb.Name,
}
}
case "content_block_delta":
delta := event.Delta
switch delta.Type {
case "text_delta":
textContent.WriteString(delta.Text)
if onChunk != nil {
onChunk(textContent.String())
}
case "input_json_delta":
if tb, ok := activeTools[event.Index]; ok {
tb.inputJSON.WriteString(delta.PartialJSON)
}
}
case "content_block_stop":
if tb, ok := activeTools[event.Index]; ok {
var args map[string]any
if err := json.Unmarshal([]byte(tb.inputJSON.String()), &args); err != nil {
log.Printf("anthropic stream: failed to decode tool call input for %q: %v", tb.name, err)
args = map[string]any{"raw": tb.inputJSON.String()}
}
toolCalls = append(toolCalls, ToolCall{
ID: tb.id,
Name: tb.name,
Arguments: args,
})
delete(activeTools, event.Index)
}
case "message_delta":
stopReason = event.Delta.StopReason
if event.Usage.OutputTokens > 0 {
outputTokens = event.Usage.OutputTokens
}
}
}
if err := stream.Err(); err != nil {
return nil, fmt.Errorf("claude streaming API call: %w", err)
}
finishReason := "stop"
switch stopReason {
case anthropic.StopReasonToolUse:
finishReason = "tool_calls"
case anthropic.StopReasonMaxTokens:
finishReason = "length"
case anthropic.StopReasonEndTurn:
finishReason = "stop"
}
return &LLMResponse{
Content: textContent.String(),
ToolCalls: toolCalls,
FinishReason: finishReason,
Usage: &UsageInfo{
PromptTokens: int(inputTokens),
CompletionTokens: int(outputTokens),
TotalTokens: int(inputTokens + outputTokens),
},
}, nil
}
func (p *Provider) GetDefaultModel() string {
return "claude-sonnet-4.6"
}
func (p *Provider) BaseURL() string {
return p.baseURL
}
func buildParams(
messages []Message,
tools []ToolDefinition,
model string,
options map[string]any,
) (anthropic.MessageNewParams, error) {
var system []anthropic.TextBlockParam
var anthropicMessages []anthropic.MessageParam
for _, msg := range messages {
switch msg.Role {
case "system":
// Prefer structured SystemParts for per-block cache_control.
// This enables LLM-side KV cache reuse: the static block's prefix
// hash stays stable across requests while dynamic parts change freely.
if len(msg.SystemParts) > 0 {
for _, part := range msg.SystemParts {
block := anthropic.TextBlockParam{Text: part.Text}
if part.CacheControl != nil && part.CacheControl.Type == "ephemeral" {
block.CacheControl = anthropic.NewCacheControlEphemeralParam()
}
system = append(system, block)
}
} else {
system = append(system, anthropic.TextBlockParam{Text: msg.Content})
}
case "user":
if msg.ToolCallID != "" {
anthropicMessages = append(anthropicMessages,
anthropic.NewUserMessage(anthropic.NewToolResultBlock(msg.ToolCallID, msg.Content, false)),
)
} else {
anthropicMessages = append(anthropicMessages,
anthropic.NewUserMessage(anthropic.NewTextBlock(msg.Content)),
)
}
case "assistant":
if len(msg.ToolCalls) > 0 {
var blocks []anthropic.ContentBlockParamUnion
if msg.Content != "" {
blocks = append(blocks, anthropic.NewTextBlock(msg.Content))
}
for _, tc := range msg.ToolCalls {
blocks = append(blocks, anthropic.NewToolUseBlock(tc.ID, tc.Arguments, tc.Name))
}
anthropicMessages = append(anthropicMessages, anthropic.NewAssistantMessage(blocks...))
} else {
anthropicMessages = append(anthropicMessages,
anthropic.NewAssistantMessage(anthropic.NewTextBlock(msg.Content)),
)
}
case "tool":
anthropicMessages = append(anthropicMessages,
anthropic.NewUserMessage(anthropic.NewToolResultBlock(msg.ToolCallID, msg.Content, false)),
)
}
}
maxTokens := int64(4096)
if mt, ok := options["max_tokens"].(int); ok {
maxTokens = int64(mt)
}
params := anthropic.MessageNewParams{
Model: anthropic.Model(model),
Messages: anthropicMessages,
MaxTokens: maxTokens,
}
if len(system) > 0 {
params.System = system
}
if temp, ok := options["temperature"].(float64); ok {
params.Temperature = anthropic.Float(temp)
}
if len(tools) > 0 {
params.Tools = translateTools(tools)
}
return params, nil
}
func translateTools(tools []ToolDefinition) []anthropic.ToolUnionParam {
result := make([]anthropic.ToolUnionParam, 0, len(tools))
for _, t := range tools {
tool := anthropic.ToolParam{
Name: t.Function.Name,
InputSchema: anthropic.ToolInputSchemaParam{
Properties: t.Function.Parameters["properties"],
},
}
if desc := t.Function.Description; desc != "" {
tool.Description = anthropic.String(desc)
}
if req, ok := t.Function.Parameters["required"].([]any); ok {
required := make([]string, 0, len(req))
for _, r := range req {
if s, ok := r.(string); ok {
required = append(required, s)
}
}
tool.InputSchema.Required = required
}
result = append(result, anthropic.ToolUnionParam{OfTool: &tool})
}
return result
}
func parseResponse(resp *anthropic.Message) *LLMResponse {
var content strings.Builder
var toolCalls []ToolCall
for _, block := range resp.Content {
switch block.Type {
case "text":
tb := block.AsText()
content.WriteString(tb.Text)
case "tool_use":
tu := block.AsToolUse()
var args map[string]any
if err := json.Unmarshal(tu.Input, &args); err != nil {
log.Printf("anthropic: failed to decode tool call input for %q: %v", tu.Name, err)
args = map[string]any{"raw": string(tu.Input)}
}
toolCalls = append(toolCalls, ToolCall{
ID: tu.ID,
Name: tu.Name,
Arguments: args,
})
}
}
finishReason := "stop"
switch resp.StopReason {
case anthropic.StopReasonToolUse:
finishReason = "tool_calls"
case anthropic.StopReasonMaxTokens:
finishReason = "length"
case anthropic.StopReasonEndTurn:
finishReason = "stop"
}
return &LLMResponse{
Content: content.String(),
ToolCalls: toolCalls,
FinishReason: finishReason,
Usage: &UsageInfo{
PromptTokens: int(resp.Usage.InputTokens),
CompletionTokens: int(resp.Usage.OutputTokens),
TotalTokens: int(resp.Usage.InputTokens + resp.Usage.OutputTokens),
},
}
}
func normalizeBaseURL(apiBase string) string {
base := strings.TrimSpace(apiBase)
if base == "" {
return defaultBaseURL
}
base = strings.TrimRight(base, "/")
if before, ok := strings.CutSuffix(base, "/v1"); ok {
base = before
}
if base == "" {
return defaultBaseURL
}
return base
}