golangLAKEHOUSE/internal/chat/openai_compat.go
root 0efc7363c5 scrum 2026-04-30: 4 real fixes + 2 INFOs from cross-lineage review
3-lineage scrum (Opus 4.7 / Kimi K2.6 / Qwen3-coder) on today's wave
landed 4 real findings (2 BLOCK + 2 WARN) and 2 INFO touch-ups.
Verbatim verdicts + disposition table at:
  reports/scrum/_evidence/2026-04-30/

B-1 (BLOCK Opus + INFO Kimi convergent) — ResolveKey API:
  collapse from 3-arg (envVar, envFileName, envFilePath) to 2-arg
  (envVar, envFilePath). Pre-fix every chatd caller passed the env
  var name twice; if operator renamed *_key_env in lakehouse.toml
  while keeping the canonical KEY= line in the .env file, fallback
  silently missed.

B-2 (WARN Opus + WARN Kimi convergent) — handleProviders probe:
  drop the synthesize-then-Resolve probe; look up by name directly
  via Registry.Available(name). Prior probe synthesized "<name>/probe"
  model strings and routed through Resolve, fragile to any future
  routing rule (e.g. cloud-suffix special case).

B-3 (BLOCK Opus single — verified by trace + end-to-end probe) —
  OllamaCloud.Chat StripPrefix used "cloud" but registry routes
  "ollama_cloud/<m>". Result: upstream got the prefixed model name
  and 400'd. Smoke missed it because chatd_smoke runs without
  ollama_cloud registered. Now strips the right prefix; new
  TestOllamaCloud_StripsCorrectPrefix locks both prefix + suffix
  cases. Verified live: ollama_cloud/deepseek-v3.2 round-trips
  cleanly through the real ollama.com endpoint.

B-4 (WARN Opus single) — Ollama finishReason: read done_reason
  field instead of inferring from done bool alone. Newer Ollama
  reports done=true with done_reason="length" on truncation; the
  prior code mapped that to "stop" and lost the truncation signal
  the playbook_lift judge needs to retry. New
  TestFinishReasonFromOllama_PrefersDoneReason covers the fallback
  ladder.

INFOs:
- B-5: replace hand-rolled insertion sort in Registry.Names with
  sort.Strings (Opus called the "avoid sort import" comment a
  false economy — correct).
- A-1: clarify the playbook_lift.sh comment around -judge "" arg
  passing (Opus noted the comment said "env priority" but didn't
  reflect that the empty arg also passes through the Go driver's
  resolution chain).

False positives dismissed (3, documented in disposition.md):
- Kimi: TestMaybeDowngrade_WithConfigList wrong assertion (test IS
  correct per design — model excluded from weak list = strong = downgrade)
- Qwen: nil-deref claim (defensive code already handles nil)
- Opus: qwen3.5:latest doesn't exist on Ollama hub (true on the
  public hub but local install has it)

just verify: PASS. chatd_smoke 6/6 PASS. New regression tests:
3 (B-2, B-3, B-4 each get a focused test).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-30 00:28:08 -05:00

150 lines
4.9 KiB
Go

package chat
import (
"bytes"
"context"
"encoding/json"
"errors"
"fmt"
"io"
"net/http"
"strings"
"time"
)
// openaiCompat implements the OpenAI Chat Completions wire format,
// which OpenRouter / OpenCode / Kimi all speak. Differences between
// those three providers are limited to:
// - base URL (e.g. /v1/chat/completions vs /zen/v1/chat/completions)
// - prefix stripped from req.Model
// - bearer token source (different env vars)
//
// Each concrete provider wraps this with its own constructor pinning
// those values.
type openaiCompat struct {
name string
baseURL string
apiKey string
prefix string
httpClient *http.Client
}
func newOpenAICompat(name, baseURL, apiKey, prefix string, timeout time.Duration) *openaiCompat {
if timeout == 0 {
timeout = 180 * time.Second
}
return &openaiCompat{
name: name,
baseURL: strings.TrimRight(baseURL, "/"),
apiKey: apiKey,
prefix: prefix,
httpClient: &http.Client{Timeout: timeout},
}
}
func (c *openaiCompat) Name() string { return c.name }
func (c *openaiCompat) Available() bool { return c.apiKey != "" }
func (c *openaiCompat) Chat(ctx context.Context, req Request) (*Response, error) {
model := StripPrefix(req.Model, c.prefix)
body := map[string]any{
"model": model,
"messages": req.Messages,
"stream": false,
}
// Anthropic 4.7 (via OpenCode → Anthropic) rejects the temperature
// field entirely with "temperature is deprecated for this model".
// Send only when caller explicitly set it. Nil = use upstream
// default. Keep this true for ALL openai-compat providers — being
// permissive about field omission is better than maintaining a
// per-model deprecation table.
if req.Temperature != nil {
body["temperature"] = *req.Temperature
}
if req.MaxTokens > 0 {
body["max_tokens"] = req.MaxTokens
}
if req.Format == "json" {
// OpenAI-compat response_format. OpenRouter passes through;
// OpenCode + Kimi accept it. Worst case the upstream ignores it.
body["response_format"] = map[string]any{"type": "json_object"}
}
bs, _ := json.Marshal(body)
url := c.baseURL + "/chat/completions"
httpReq, err := http.NewRequestWithContext(ctx, "POST", url, bytes.NewReader(bs))
if err != nil {
return nil, err
}
httpReq.Header.Set("Content-Type", "application/json")
httpReq.Header.Set("Authorization", "Bearer "+c.apiKey)
// HTTP-Referer + X-Title are OpenRouter-specific but harmless on
// other providers — they pass them through to upstream attribution.
httpReq.Header.Set("HTTP-Referer", "https://golanglakehouse.local")
httpReq.Header.Set("X-Title", "Lakehouse-Go")
resp, err := c.httpClient.Do(httpReq)
if err != nil {
if errors.Is(ctx.Err(), context.DeadlineExceeded) {
return nil, fmt.Errorf("%w: %s", ErrTimeout, c.name)
}
return nil, fmt.Errorf("%s: %w", c.name, err)
}
defer resp.Body.Close()
rb, _ := io.ReadAll(resp.Body)
if resp.StatusCode/100 != 2 {
return nil, fmt.Errorf("%w: %s %d: %s", ErrUpstream, c.name, resp.StatusCode, abbrev(string(rb), 300))
}
var oaResp struct {
Model string `json:"model"`
Choices []struct {
Message struct {
Content string `json:"content"`
} `json:"message"`
FinishReason string `json:"finish_reason"`
} `json:"choices"`
Usage struct {
PromptTokens int `json:"prompt_tokens"`
CompletionTokens int `json:"completion_tokens"`
} `json:"usage"`
}
if err := json.Unmarshal(rb, &oaResp); err != nil {
return nil, fmt.Errorf("%s decode: %w (body=%s)", c.name, err, abbrev(string(rb), 200))
}
if len(oaResp.Choices) == 0 {
return nil, fmt.Errorf("%w: %s returned 0 choices: %s", ErrUpstream, c.name, abbrev(string(rb), 200))
}
return &Response{
Model: model,
Content: oaResp.Choices[0].Message.Content,
InputTokens: oaResp.Usage.PromptTokens,
OutputTokens: oaResp.Usage.CompletionTokens,
FinishReason: oaResp.Choices[0].FinishReason,
}, nil
}
// NewOpenRouter returns a provider for openrouter.ai. apiKey
// resolved at construction (typically from OPENROUTER_API_KEY env or
// /etc/lakehouse/openrouter.env).
func NewOpenRouter(apiKey string, timeout time.Duration) *openaiCompat {
return newOpenAICompat("openrouter", "https://openrouter.ai/api/v1", apiKey, "openrouter", timeout)
}
// NewOpenCode returns a provider for OpenCode's unified Zen+Go endpoint.
// One key reaches Anthropic Opus, GPT-5, Gemini 3.1, Kimi K2.6,
// DeepSeek, GLM, Qwen, plus 4 free-tier models.
func NewOpenCode(apiKey string, timeout time.Duration) *openaiCompat {
return newOpenAICompat("opencode", "https://opencode.ai/zen/v1", apiKey, "opencode", timeout)
}
// NewKimi returns a provider for the direct Kimi For Coding endpoint.
// `api.kimi.com` is a separate account system from api.moonshot.ai —
// keys are NOT interchangeable.
func NewKimi(apiKey string, timeout time.Duration) *openaiCompat {
return newOpenAICompat("kimi", "https://api.kimi.com/coding/v1", apiKey, "kimi", timeout)
}