Automated salvage commit — agent session ended (exit 124). Work in progress, may need continuation.
176 lines
6.0 KiB
Python
176 lines
6.0 KiB
Python
"""Paperclip integration for Timmy.
|
|
|
|
This module provides a client for the Paperclip API, and a poller for
|
|
running research tasks.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import asyncio
|
|
import logging
|
|
from dataclasses import dataclass
|
|
|
|
import httpx
|
|
|
|
from config import settings
|
|
from timmy.research_tools import get_llm_client, google_web_search
|
|
from timmy.research_triage import triage_research_report
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
|
|
@dataclass
|
|
class PaperclipTask:
|
|
"""A task from the Paperclip API."""
|
|
|
|
id: str
|
|
kind: str
|
|
context: dict
|
|
|
|
|
|
class PaperclipClient:
|
|
"""A client for the Paperclip API."""
|
|
|
|
def __init__(self) -> None:
|
|
self.base_url = settings.paperclip_url
|
|
self.api_key = settings.paperclip_api_key
|
|
self.agent_id = settings.paperclip_agent_id
|
|
self.company_id = settings.paperclip_company_id
|
|
self.timeout = settings.paperclip_timeout
|
|
|
|
async def get_tasks(self) -> list[PaperclipTask]:
|
|
"""Get a list of tasks from the Paperclip API."""
|
|
async with httpx.AsyncClient(timeout=self.timeout) as client:
|
|
resp = await client.get(
|
|
f"{self.base_url}/api/tasks",
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
params={
|
|
"agent_id": self.agent_id,
|
|
"company_id": self.company_id,
|
|
"status": "queued",
|
|
},
|
|
)
|
|
resp.raise_for_status()
|
|
tasks = resp.json()
|
|
return [
|
|
PaperclipTask(id=t["id"], kind=t["kind"], context=t["context"])
|
|
for t in tasks
|
|
]
|
|
|
|
async def update_task_status(
|
|
self, task_id: str, status: str, result: str | None = None
|
|
) -> None:
|
|
"""Update the status of a task."""
|
|
async with httpx.AsyncClient(timeout=self.timeout) as client:
|
|
await client.patch(
|
|
f"{self.base_url}/api/tasks/{task_id}",
|
|
headers={"Authorization": f"Bearer {self.api_key}"},
|
|
json={"status": status, "result": result},
|
|
)
|
|
|
|
|
|
class ResearchOrchestrator:
|
|
"""Orchestrates research tasks."""
|
|
|
|
async def get_gitea_issue(self, issue_number: int) -> dict:
|
|
"""Get a Gitea issue by its number."""
|
|
owner, repo = settings.gitea_repo.split("/", 1)
|
|
api_url = f"{settings.gitea_url}/api/v1/repos/{owner}/{repo}/issues/{issue_number}"
|
|
async with httpx.AsyncClient(timeout=15) as client:
|
|
resp = await client.get(
|
|
api_url,
|
|
headers={"Authorization": f"token {settings.gitea_token}"},
|
|
)
|
|
resp.raise_for_status()
|
|
return resp.json()
|
|
|
|
async def post_gitea_comment(self, issue_number: int, comment: str) -> None:
|
|
"""Post a comment to a Gitea issue."""
|
|
owner, repo = settings.gitea_repo.split("/", 1)
|
|
api_url = f"{settings.gitea_url}/api/v1/repos/{owner}/{repo}/issues/{issue_number}/comments"
|
|
async with httpx.AsyncClient(timeout=15) as client:
|
|
await client.post(
|
|
api_url,
|
|
headers={"Authorization": f"token {settings.gitea_token}"},
|
|
json={"body": comment},
|
|
)
|
|
|
|
async def run_research_pipeline(self, issue_title: str) -> str:
|
|
"""Run the research pipeline."""
|
|
search_results = await google_web_search(issue_title)
|
|
|
|
llm_client = get_llm_client()
|
|
response = await llm_client.completion(
|
|
f"Summarize the following search results and generate a research report:\\n\\n{search_results}",
|
|
max_tokens=2048,
|
|
)
|
|
return response.text
|
|
|
|
async def run(self, context: dict) -> str:
|
|
"""Run a research task."""
|
|
issue_number = context.get("issue_number")
|
|
if not issue_number:
|
|
return "Missing issue_number in task context"
|
|
|
|
issue = await self.get_gitea_issue(issue_number)
|
|
|
|
report = await self.run_research_pipeline(issue["title"])
|
|
|
|
triage_results = await triage_research_report(report, source_issue=issue_number)
|
|
|
|
comment = f"Research complete for issue #{issue_number}.\\n\\n"
|
|
if triage_results:
|
|
comment += "Created the following issues:\\n"
|
|
for result in triage_results:
|
|
if result["gitea_issue"]:
|
|
comment += f"- #{result['gitea_issue']['number']}: {result['action_item'].title}\\n"
|
|
else:
|
|
comment += "No new issues were created.\\n"
|
|
|
|
await self.post_gitea_comment(issue_number, comment)
|
|
|
|
return f"Research complete for issue #{issue_number}"
|
|
|
|
|
|
class PaperclipPoller:
|
|
"""Polls the Paperclip API for new tasks."""
|
|
|
|
def __init__(self) -> None:
|
|
self.client = PaperclipClient()
|
|
self.orchestrator = ResearchOrchestrator()
|
|
self.poll_interval = settings.paperclip_poll_interval
|
|
|
|
async def poll(self) -> None:
|
|
"""Poll the Paperclip API for new tasks."""
|
|
if self.poll_interval == 0:
|
|
return
|
|
|
|
while True:
|
|
try:
|
|
tasks = await self.client.get_tasks()
|
|
for task in tasks:
|
|
if task.kind == "research":
|
|
await self.run_research_task(task)
|
|
except httpx.HTTPError as exc:
|
|
logger.warning("Error polling Paperclip: %s", exc)
|
|
|
|
await asyncio.sleep(self.poll_interval)
|
|
|
|
async def run_research_task(self, task: PaperclipTask) -> None:
|
|
"""Run a research task."""
|
|
await self.client.update_task_status(task.id, "running")
|
|
try:
|
|
result = await self.orchestrator.run(task.context)
|
|
await self.client.update_task_status(task.id, "completed", result)
|
|
except Exception as exc:
|
|
logger.error("Error running research task: %s", exc, exc_info=True)
|
|
await self.client.update_task_status(task.id, "failed", str(exc))
|
|
|
|
|
|
async def start_paperclip_poller() -> None:
|
|
"""Start the Paperclip poller."""
|
|
if settings.paperclip_enabled:
|
|
poller = PaperclipPoller()
|
|
asyncio.create_task(poller.poll())
|
|
|