正确转发 Usage 数据#2
Open
kazaff wants to merge 3 commits into
Open
Conversation
feat(proxy): add TypeScript implementation of DeepSeek proxy server Add a complete TypeScript proxy server that converts OpenAI Responses API requests to DeepSeek Chat Completions API. This includes support for streaming/non-streaming responses, tool/function call conversions, and XML function call parsing. - Convert OpenAI Responses API requests to DeepSeek Chat Completions format - Support both streaming and non-streaming responses - Handle tool/function call conversions - Parse XML function calls from DeepSeek responses - Map model names between OpenAI and DeepSeek - Provide health check endpoint - Implement graceful shutdown with signal handling Add AGENTS.md documentation covering project overview, features, API endpoints, environment variables, and development instructions. chore(converter): add usage conversion function for DeepSeek responses Add convert_deepseek_usage_to_responses function to properly map DeepSeek usage metrics (prompt_tokens, completion_tokens, total_tokens) to the expected response format. Apply this conversion across all response types including tool calls and regular responses. ```
Review Summary by QodoFix usage data conversion and forwarding in DeepSeek proxy
WalkthroughsDescription• Add convert_deepseek_usage_to_responses() function to properly map DeepSeek usage metrics to OpenAI Responses API format • Apply usage conversion across all response types (tool calls, regular, XML-based) • Pass usage data from DeepSeek stream chunks to StreamChunk output • Add comprehensive AGENTS.md documentation for TypeScript proxy server Diagramflowchart LR
A["DeepSeek Response<br/>with usage"] -->|convert_deepseek_usage_to_responses| B["Mapped Usage<br/>prompt_tokens → input_tokens"]
B -->|apply to all responses| C["Tool Calls Response"]
B -->|apply to all responses| D["Regular Response"]
B -->|apply to all responses| E["XML Tool Calls"]
B -->|apply to stream chunks| F["Stream Chunk Output"]
C -->|OpenAI Responses API| G["Downstream Apps"]
D -->|OpenAI Responses API| G
E -->|OpenAI Responses API| G
F -->|OpenAI Responses API| G
File Changes1. src/converter.ts
|
Code Review by Qodo
1. convert_deepseek_usage_to_responses uses any
|
![qodo-code-review[bot]](https://avatars.githubusercontent.com/in/484649?s=60&v=4){kind=small}
Comment on lines
+727
to
+736
| export function convert_deepseek_usage_to_responses(usage: any): any { | ||
| if (!usage || typeof usage !== 'object') { | ||
| return { input_tokens: 0, output_tokens: 0, total_tokens: 0 }; | ||
| } | ||
| return { | ||
| input_tokens: usage.prompt_tokens ?? 0, | ||
| output_tokens: usage.completion_tokens ?? 0, | ||
| total_tokens: usage.total_tokens ?? 0 | ||
| }; | ||
| } |
There was a problem hiding this comment.
1. convert_deepseek_usage_to_responses uses any 📘 Rule violation ⚙ Maintainability
The newly added convert_deepseek_usage_to_responses(usage: any): any bypasses strict TypeScript typing by introducing any in both parameter and return types. This weakens type safety and can mask schema/field mismatches at compile time.
Agent Prompt
## Issue description
New code introduces `any` types (`usage: any` and return `any`) in `convert_deepseek_usage_to_responses`, which weakens strict typing guarantees.
## Issue Context
The project compliance requires strict TypeScript typing; this helper can be typed with a small interface/type describing DeepSeek usage input and the normalized Responses usage output.
## Fix Focus Areas
- src/converter.ts[727-736]
ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools
Comment on lines
+727
to
+735
| export function convert_deepseek_usage_to_responses(usage: any): any { | ||
| if (!usage || typeof usage !== 'object') { | ||
| return { input_tokens: 0, output_tokens: 0, total_tokens: 0 }; | ||
| } | ||
| return { | ||
| input_tokens: usage.prompt_tokens ?? 0, | ||
| output_tokens: usage.completion_tokens ?? 0, | ||
| total_tokens: usage.total_tokens ?? 0 | ||
| }; |
There was a problem hiding this comment.
2. Sse usage变成0 🐞 Bug ≡ Correctness
convert_deepseek_usage_to_responses 将 usage 字段从 DeepSeek 的 prompt_tokens/completion_tokens 改为 input_tokens/output_tokens，但 streaming.ts 的 build_completed_event 仍只读取 prompt_tokens/completion_tokens。结果是在“非流式响应被包装成模拟 SSE”时，response.completed 会输出全 0 的 usage。
Agent Prompt
## Issue description
非流式转换后的 `responses_response.usage` 现在是 `{input_tokens, output_tokens, total_tokens}`，但 `streaming.ts` 的 `build_completed_event()` 仍按 `{prompt_tokens, completion_tokens, total_tokens}` 映射，导致模拟 SSE 的 `response.completed` usage 归零。
## Issue Context
- `convert_deepseek_usage_to_responses()` 将 DeepSeek usage 键名改成 Responses usage。
- `handle_non_streaming_response()` 在 `original_request.stream=true` 时会调用 `generate_simulated_sse(responses_response, ...)`。
- `generate_simulated_sse()` 把 `responses_response.usage` 传给 `build_completed_event()`，但后者只读取 `prompt_tokens/completion_tokens`。
## Fix Focus Areas
- src/streaming.ts[596-612]
- src/streaming.ts[614-708]
## Suggested fix
在 `build_completed_event(response_id, usage)` 中同时兼容两套字段：
- 若存在 `usage.input_tokens`/`usage.output_tokens`，直接使用它们
- 否则回退到 `usage.prompt_tokens`/`usage.completion_tokens`
并保证 `total_tokens` 也按同样逻辑取值。这样模拟 SSE 与真实 streaming 两条路径都能得到正确 usage。
ⓘ Copy this prompt and use it to remediate the issue with your preferred AI generation tools
feat(security): sanitize sensitive data from logs - Disable full request logging in converter.ts that could contain sensitive data - Implement header sanitization in server.ts to redact authorization, x-api-key, and shutdown secrets with truncated values - Remove debug file writing functionality that could leak sensitive data - Disable message logging in streaming.ts that could expose sensitive information ```
xiaoshaoning
force-pushed
the
main
branch
2 times, most recently
from
9a8b70b to
7738511
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
用于让 CC-Switch 类的下游应用可以拿到完整正确的响应结果。