phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-10-11 05:38:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

feat(tests): make inference_recorder into api_recorder (include tool_invoke) (#3403)

Browse source 
Renames `inference_recorder.py` to `api_recorder.py` and extends it to
support recording/replaying tool invocations in addition to inference
calls.

This allows us to record web-search, etc. tool calls and thereafter
apply recordings for `tests/integration/responses`

## Test Plan

```
export OPENAI_API_KEY=...
export TAVILY_SEARCH_API_KEY=...

./scripts/integration-tests.sh --stack-config ci-tests \
   --suite responses --inference-mode record-if-missing
```
This commit is contained in:
Ashwin Bharambe 2025-10-09 14:27:51 -07:00 • committed by GitHub
parent 26fd5dbd34
commit f50ce11a3b
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
284 changed files with 296191 additions and 631 deletions
Download patch file Download diff file Expand all files Collapse all files

34
tests/common/mcp.py
View file

@ -159,7 +159,6 @@ def make_mcp_server(required_auth_token: str | None = None, tools: dict[str, Cal
import threading
import time
import httpx
import uvicorn
from mcp.server.fastmcp import FastMCP
from mcp.server.sse import SseServerTransport
@ -171,6 +170,11 @@ def make_mcp_server(required_auth_token: str | None = None, tools: dict[str, Cal
server = FastMCP("FastMCP Test Server", log_level="WARNING")
# Silence verbose MCP server logs
import logging # allow-direct-logging
logging.getLogger("mcp.server.lowlevel.server").setLevel(logging.WARNING)
tools = tools or default_tools()
# Register all tools with the server
@ -234,29 +238,25 @@ def make_mcp_server(required_auth_token: str | None = None, tools: dict[str, Cal
logger.debug(f"Starting MCP server thread on port {port}")
server_thread.start()
# Polling until the server is ready
timeout = 10
# Wait for the server thread to be running
# Note: We can't use a simple HTTP GET health check on /sse because it's an SSE endpoint
# that expects a long-lived connection, not a simple request/response
timeout = 2
start_time = time.time()
server_url = f"http://localhost:{port}/sse"
logger.debug(f"Waiting for MCP server to be ready at {server_url}")
logger.debug(f"Waiting for MCP server thread to start on port {port}")
while time.time() - start_time < timeout:
try:
response = httpx.get(server_url)
if response.status_code in [200, 401]:
logger.debug(f"MCP server is ready on port {port} (status: {response.status_code})")
break
except httpx.RequestError as e:
logger.debug(f"Server not ready yet, retrying... ({e})")
pass
time.sleep(0.1)
if server_thread.is_alive():
# Give the server a moment to bind to the port
time.sleep(0.1)
logger.debug(f"MCP server is ready on port {port}")
break
time.sleep(0.05)
else:
# If we exit the loop due to timeout
logger.error(f"MCP server failed to start within {timeout} seconds on port {port}")
logger.error(f"Thread alive: {server_thread.is_alive()}")
if server_thread.is_alive():
logger.error("Server thread is still running but not responding to HTTP requests")
logger.error(f"MCP server thread failed to start within {timeout} seconds on port {port}")
try:
yield {"server_url": server_url}