1
0

[gemini] feat: Integrate ResearchOrchestrator with Paperclip (#978) (#1111)

Co-authored-by: Google Gemini <gemini@hermes.local>
Co-committed-by: Google Gemini <gemini@hermes.local>
This commit is contained in:
2026-03-23 18:09:29 +00:00
committed by rockachopa
parent 19dbdec314
commit 852fec3681
5 changed files with 922 additions and 33 deletions

726
poetry.lock generated

File diff suppressed because it is too large Load Diff

View File

@@ -68,7 +68,7 @@ voice = ["pyttsx3", "openai-whisper", "piper-tts", "sounddevice"]
celery = ["celery"]
embeddings = ["sentence-transformers", "numpy"]
git = ["GitPython"]
research = ["requests", "trafilatura"]
research = ["requests", "trafilatura", "google-search-results"]
dev = ["pytest", "pytest-asyncio", "pytest-cov", "pytest-timeout", "pytest-randomly", "pytest-xdist", "selenium"]
[tool.poetry.group.dev.dependencies]

View File

@@ -375,13 +375,21 @@ def _startup_init() -> None:
def _startup_background_tasks() -> list[asyncio.Task]:
"""Spawn all recurring background tasks (non-blocking)."""
return [
bg_tasks = [
asyncio.create_task(_briefing_scheduler()),
asyncio.create_task(_thinking_scheduler()),
asyncio.create_task(_loop_qa_scheduler()),
asyncio.create_task(_presence_watcher()),
asyncio.create_task(_start_chat_integrations_background()),
]
try:
from timmy.paperclip import start_paperclip_poller
bg_tasks.append(asyncio.create_task(start_paperclip_poller()))
logger.info("Paperclip poller started")
except ImportError:
logger.debug("Paperclip module not found, skipping poller")
return bg_tasks
def _try_prune(label: str, prune_fn, days: int) -> None:

175
src/timmy/paperclip.py Normal file
View File

@@ -0,0 +1,175 @@
"""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_triage import triage_research_report
from timmy.research_tools import google_web_search, get_llm_client
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())

View File

@@ -0,0 +1,42 @@
"""Tools for the research pipeline."""
from __future__ import annotations
import logging
import os
from typing import Any
from config import settings
from serpapi import GoogleSearch
logger = logging.getLogger(__name__)
async def google_web_search(query: str) -> str:
"""Perform a Google search and return the results."""
if "SERPAPI_API_KEY" not in os.environ:
logger.warning("SERPAPI_API_KEY not set, skipping web search")
return ""
params = {
"q": query,
"api_key": os.environ["SERPAPI_API_KEY"],
}
search = GoogleSearch(params)
results = search.get_dict()
return str(results)
def get_llm_client() -> Any:
"""Get an LLM client."""
# This is a placeholder. In a real application, this would return
# a client for an LLM service like OpenAI, Anthropic, or a local
# model.
class MockLLMClient:
async def completion(self, prompt: str, max_tokens: int) -> Any:
class MockCompletion:
def __init__(self, text: str) -> None:
self.text = text
return MockCompletion(f"This is a summary of the search results for '{prompt}'.")
return MockLLMClient()