iop/apps/edge/internal/openai/chat_handler.go
toki 672b4cdb5d refactor(bridge): 실행 경계 선행 안정화를 반영한다
원격 터미널 브리지 POC 전에 Client HTTP lifecycle과 Edge run result surface 계약을 고정해야 하므로 관련 구현, 테스트, 로드맵 상태를 함께 정리한다.
2026-06-07 08:40:36 +09:00

213 lines
6 KiB
Go

package openai
import (
"context"
"encoding/json"
"errors"
"fmt"
"net/http"
"strings"
"time"
"go.uber.org/zap"
edgeservice "iop/apps/edge/internal/service"
)
func (s *Server) handleChatCompletions(w http.ResponseWriter, r *http.Request) {
if r.Method != http.MethodPost {
writeError(w, http.StatusMethodNotAllowed, "method_not_allowed", "method not allowed")
return
}
defer r.Body.Close()
var req chatCompletionRequest
if err := json.NewDecoder(r.Body).Decode(&req); err != nil {
writeError(w, http.StatusBadRequest, "invalid_request_error", "invalid JSON request")
return
}
target := s.resolveTarget(req.Model)
if target == "" {
writeError(w, http.StatusBadRequest, "invalid_request_error", "model is required")
return
}
basePrompt := promptFromMessages(req.Messages)
if strings.TrimSpace(basePrompt) == "" {
writeError(w, http.StatusBadRequest, "invalid_request_error", "messages are required")
return
}
outputPolicy := s.resolveOutputPolicy(basePrompt)
messages := req.Messages
if instruction := strictOutputContractInstruction(outputPolicy); instruction != "" {
messages = prependSystemMessage(messages, instruction)
}
prompt := promptFromMessages(messages)
input := req.runInput(prompt, messages, outputPolicy.Strict)
s.logger.Info("openai chat completion input",
zap.String("model", req.Model),
zap.String("target", target),
zap.String("adapter", s.resolveAdapter()),
zap.Bool("strict_output", outputPolicy.Strict),
zap.Bool("strict_stream_buffer", outputPolicy.StreamBuffer),
zap.String("xml_completion_tool", outputPolicy.XMLCompletionTool),
zap.Bool("contract_instruction", outputPolicy.ContractInstruction),
zap.Bool("stream", req.Stream),
zap.Int("message_count", len(req.Messages)),
zap.Int("prompt_len", len(prompt)),
zap.String("prompt_preview", previewString(prompt, 1000)),
zap.Any("input_keys", mapKeys(input)),
)
handle, err := s.service.SubmitRun(r.Context(), edgeservice.SubmitRunRequest{
NodeRef: s.cfg.NodeRef,
Adapter: s.resolveAdapter(),
Target: target,
SessionID: s.resolveSessionID(),
Prompt: prompt,
Input: input,
TimeoutSec: s.resolveTimeoutSec(),
Metadata: map[string]string{
"source": "openai",
"openai_model": req.Model,
"openai_stream": fmt.Sprintf("%t", req.Stream),
"strict_output": fmt.Sprintf("%t", outputPolicy.Strict),
},
})
if err != nil {
writeError(w, http.StatusBadGateway, "node_dispatch_error", err.Error())
return
}
defer handle.Close()
if req.Stream {
s.streamChatCompletion(w, r, req, handle, outputPolicy)
return
}
s.completeChatCompletion(w, r, req, handle, outputPolicy)
}
func (s *Server) completeChatCompletion(w http.ResponseWriter, r *http.Request, req chatCompletionRequest, handle edgeservice.RunResult, outputPolicy strictOutputPolicy) {
text, reasoning, usage, err := collectRunResult(r.Context(), handle.Stream(), handle.WaitTimeout())
if err != nil {
writeError(w, httpStatusForRunError(err), "run_error", err.Error())
return
}
text, reasoning, normalized := normalizeCompletionOutput(outputPolicy, text, reasoning)
s.logger.Info("openai chat completion output",
zap.String("run_id", handle.Dispatch().RunID),
zap.Bool("strict_output", outputPolicy.Strict),
zap.Bool("strict_stream_buffer", outputPolicy.StreamBuffer),
zap.String("xml_completion_tool", outputPolicy.XMLCompletionTool),
zap.Bool("normalized", normalized),
zap.Int("content_len", len(text)),
zap.Int("reasoning_len", len(reasoning)),
zap.String("content_preview", previewString(text, 1000)),
)
created := time.Now().Unix()
writeJSON(w, http.StatusOK, chatCompletionResponse{
ID: "chatcmpl-" + handle.Dispatch().RunID,
Object: "chat.completion",
Created: created,
Model: responseModel(req.Model, handle.Dispatch().Target),
Choices: []chatCompletionChoice{{
Index: 0,
Message: chatMessage{Role: "assistant", Content: text, ReasoningContent: reasoning},
FinishReason: "stop",
}},
Usage: usage,
})
}
func (s *Server) resolveAdapter() string {
if s.cfg.Adapter != "" {
return s.cfg.Adapter
}
return "ollama"
}
func (s *Server) resolveTarget(model string) string {
if s.cfg.Target != "" {
return s.cfg.Target
}
return strings.TrimSpace(model)
}
func (s *Server) resolveSessionID() string {
if s.cfg.SessionID != "" {
return s.cfg.SessionID
}
return edgeservice.DefaultSessionID
}
func (s *Server) resolveTimeoutSec() int {
if s.cfg.TimeoutSec > 0 {
return s.cfg.TimeoutSec
}
return edgeservice.DefaultTimeoutSec
}
func (s *Server) resolveStrictOutput() bool {
return s.cfg.StrictOutput
}
func (s *Server) resolveStrictStreamBuffer() bool {
return s.cfg.StrictStreamBuffer
}
func (s *Server) resolveOutputPolicy(prompt string) strictOutputPolicy {
policy := strictOutputPolicy{
Strict: s.resolveStrictOutput(),
StreamBuffer: s.resolveStrictStreamBuffer(),
}
if !policy.Strict {
return policy
}
policy.XMLCompletionTool, policy.XMLResultTag = inferXMLCompletionContract(prompt)
policy.ContractInstruction = policy.XMLCompletionTool != ""
return policy
}
func promptFromMessages(messages []chatMessage) string {
var b strings.Builder
for _, msg := range messages {
content := strings.TrimSpace(msg.Content)
if content == "" {
continue
}
role := strings.TrimSpace(msg.Role)
if role == "" {
role = "user"
}
if b.Len() > 0 {
b.WriteString("\n")
}
b.WriteString(role)
b.WriteString(": ")
b.WriteString(content)
}
return b.String()
}
func responseModel(requestModel, target string) string {
if requestModel != "" {
return requestModel
}
return target
}
func httpStatusForRunError(err error) int {
if errors.Is(err, context.Canceled) {
return http.StatusRequestTimeout
}
return http.StatusBadGateway
}
func writeJSON(w http.ResponseWriter, status int, v any) {
w.Header().Set("Content-Type", "application/json")
w.WriteHeader(status)
_ = json.NewEncoder(w).Encode(v)
}
func writeError(w http.ResponseWriter, status int, code, message string) {
writeJSON(w, status, errorResponse{Error: errorBody{Type: code, Message: message}})
}