feat: passthrough upstream response headers to clients

CPA previously stripped ALL response headers from upstream AI provider
APIs, preventing clients from seeing rate-limit info, request IDs,
server-timing and other useful headers.

Changes:
- Add Headers field to Response and StreamResult structs
- Add FilterUpstreamHeaders helper (hop-by-hop + security denylist)
- Add WriteUpstreamHeaders helper (respects CPA-set headers)
- ExecuteWithAuthManager/ExecuteCountWithAuthManager now return headers
- ExecuteStreamWithAuthManager returns headers from initial connection
- All 11 provider executors populate Response.Headers
- All handler call sites write filtered upstream headers before response

Filtered headers (not forwarded):
- RFC 7230 hop-by-hop: Connection, Transfer-Encoding, Keep-Alive, etc.
- Security: Set-Cookie
- CPA-managed: Content-Length, Content-Encoding
This commit is contained in:
Kirill Turanskiy
2026-02-18 00:16:22 +03:00
parent 453aaf8774
commit 1f8f198c45
24 changed files with 192 additions and 107 deletions

View File

@@ -225,7 +225,7 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
reporter.publish(ctx, parseGeminiCLIUsage(data))
var param any
out := sdktranslator.TranslateNonStream(respCtx, to, from, attemptModel, opts.OriginalRequest, payload, data, &param)
resp = cliproxyexecutor.Response{Payload: []byte(out)}
resp = cliproxyexecutor.Response{Payload: []byte(out), Headers: httpResp.Header.Clone()}
return resp, nil
}
@@ -256,7 +256,7 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
}
// ExecuteStream performs a streaming request to the Gemini CLI API.
func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (stream <-chan cliproxyexecutor.StreamChunk, err error) {
func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Auth, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (_ *cliproxyexecutor.StreamResult, err error) {
if opts.Alt == "responses/compact" {
return nil, statusErr{code: http.StatusNotImplemented, msg: "/responses/compact not supported"}
}
@@ -382,7 +382,6 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
}
out := make(chan cliproxyexecutor.StreamChunk)
stream = out
go func(resp *http.Response, reqBody []byte, attemptModel string) {
defer close(out)
defer func() {
@@ -441,7 +440,7 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
}
}(httpResp, append([]byte(nil), payload...), attemptModel)
return stream, nil
return &cliproxyexecutor.StreamResult{Headers: httpResp.Header.Clone(), Chunks: out}, nil
}
if len(lastBody) > 0 {
@@ -546,7 +545,7 @@ func (e *GeminiCLIExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.
if resp.StatusCode >= 200 && resp.StatusCode < 300 {
count := gjson.GetBytes(data, "totalTokens").Int()
translated := sdktranslator.TranslateTokenCount(respCtx, to, from, count, data)
return cliproxyexecutor.Response{Payload: []byte(translated)}, nil
return cliproxyexecutor.Response{Payload: []byte(translated), Headers: resp.Header.Clone()}, nil
}
lastStatus = resp.StatusCode
lastBody = append([]byte(nil), data...)