proxysql

Commit Graph

Author	SHA1	Message	Date
Rene Cannao	bdd2ef70e6	Add comprehensive TAP tests for GenAI async architecture This commit adds extensive TAP tests for the GenAI module's async architecture, verifying the non-blocking socketpair-based implementation. Test coverage (28 tests): - Single async request handling - Sequential embedding and rerank requests (5 each) - Batch requests (10 documents) - Mixed embedding and rerank requests - Request/response matching with varying document counts - Error handling (invalid JSON, missing fields, etc.) - Special characters (quotes, backslashes, unicode, etc.) - Large documents (5KB single, 5x1KB batch) - top_n parameter for limiting rerank results - columns parameter for result format - Concurrent connections (3 simultaneous requests) Test file: test/tap/tests/genai_async-t.cpp All tests pass successfully when GenAI module is initialized and llama-server is running at the configured endpoints. Note: These tests require: - GenAI module initialized (genai-threads > 0) - Running llama-server with embedding and rerank endpoints - Epoll support (Linux systems)	3 months ago

Author

SHA1

Message

Date

Rene Cannao

bdd2ef70e6

Add comprehensive TAP tests for GenAI async architecture

This commit adds extensive TAP tests for the GenAI module's async
architecture, verifying the non-blocking socketpair-based implementation.

Test coverage (28 tests):
- Single async request handling
- Sequential embedding and rerank requests (5 each)
- Batch requests (10 documents)
- Mixed embedding and rerank requests
- Request/response matching with varying document counts
- Error handling (invalid JSON, missing fields, etc.)
- Special characters (quotes, backslashes, unicode, etc.)
- Large documents (5KB single, 5x1KB batch)
- top_n parameter for limiting rerank results
- columns parameter for result format
- Concurrent connections (3 simultaneous requests)

Test file: test/tap/tests/genai_async-t.cpp

All tests pass successfully when GenAI module is initialized and
llama-server is running at the configured endpoints.

Note: These tests require:
- GenAI module initialized (genai-threads > 0)
- Running llama-server with embedding and rerank endpoints
- Epoll support (Linux systems)

1 Commits (ade0130e67e88b9052d0dc4e76d48d98d3328485)