agent / chat backend

pyproject.toml

@@ -5,7 +5,7 @@ description = "Add your description here"
 readme = "README.md"
 requires-python = ">=3.13"
 dependencies = [
-    "agentd>=0.1.7",
+    "agentd>=0.1.8",
     "aiofiles>=24.1.0",
     "docling>=2.41.0",
     "opensearch-py[async]>=3.0.0",

src/agent.py

@@ -0,0 +1,23 @@
+import asyncio
+
+from agentd.patch import patch_openai_with_mcp
+
+messages = [{"role": "system", "content": "You are a helpful assistant. use your tools to answer questions."}]
+
+# Async version for web server
+async def async_chat(async_client, prompt: str) -> str:
+    global messages
+    messages += [{"role": "user", "content": prompt}]
+    response = await async_client.chat.completions.create(
+        model="gpt-4.1-mini",
+        messages=messages,
+        mcp_strict=True
+    )
+
+    response_text = response.choices[0].message.content
+    print(f"user ==> {prompt}")
+    print(f"agent ==> {response_text}")
+    return response_text
+
+if __name__ == "__main__":
+    asyncio.run(async_chat("What pods are there?"))

src/app.py

@@ -2,6 +2,9 @@
 
 import os
 from collections import defaultdict
+from typing import Any
+
+from agent import async_chat
 
 os.environ['USE_CPU_ONLY'] = 'true'
 
@@ -20,7 +23,8 @@ from opensearchpy import AsyncOpenSearch
 from opensearchpy._async.http_aiohttp import AIOHttpConnection
 from docling.document_converter import DocumentConverter
 from agentd.patch import patch_openai_with_mcp
-from openai import OpenAI
+from openai import AsyncOpenAI
+from agentd.tool_decorator import tool
 
 # Initialize Docling converter
 converter = DocumentConverter()  # basic converter; tweak via PipelineOptions if you need OCR, etc.
@@ -75,7 +79,7 @@ index_body = {
     }
 }
 
-client = patch_openai_with_mcp(OpenAI())  # Get the patched client back
+async_client = patch_openai_with_mcp(AsyncOpenAI())  # Get the patched client back
 
 async def wait_for_opensearch():
     """Wait for OpenSearch to be ready with retries"""
@@ -170,7 +174,7 @@ async def process_file_common(file_path: str, file_hash: str = None):
     slim_doc = extract_relevant(full_doc)
 
     texts = [c["text"] for c in slim_doc["chunks"]]
-    resp = client.embeddings.create(model=EMBED_MODEL, input=texts)
+    resp = await async_client.embeddings.create(model=EMBED_MODEL, input=texts)
     embeddings = [d.embedding for d in resp.data]
 
     # Index each chunk as a separate document
@@ -236,15 +240,31 @@ async def upload_path(request: Request):
     return JSONResponse({"results": results})
 
 async def search(request: Request):
+
     payload = await request.json()
     query = payload.get("query")
     if not query:
         return JSONResponse({"error": "Query is required"}, status_code=400)
+    return JSONResponse(await search_tool(query))
 
+
+@tool
+async def search_tool(query: str)-> dict[str, Any]:
+    """
+    Use this tool to search for documents relevant to the query.
+
+    This endpoint accepts POST requests with a query string,
+
+    Args:
+        query (str): query string to search the corpus
+
+    Returns:
+        dict (str, Any)
+                     - {"results": [chunks]} on success
+    """
     # Embed the query
-    resp = client.embeddings.create(model=EMBED_MODEL, input=[query])
+    resp = await async_client.embeddings.create(model=EMBED_MODEL, input=[query])
     query_embedding = resp.data[0].embedding
-
     # Search using vector similarity on individual chunks
     search_body = {
         "query": {
@@ -258,9 +278,7 @@ async def search(request: Request):
         "_source": ["filename", "mimetype", "page", "text"],
         "size": 10
     }
-
     results = await es.search(index=INDEX_NAME, body=search_body)
-    
     # Transform results to match expected format
     chunks = []
     for hit in results["hits"]["hits"]:
@@ -271,13 +289,23 @@ async def search(request: Request):
             "text": hit["_source"]["text"],
             "score": hit["_score"]
         })
-    
-    return JSONResponse({"results": chunks})
+    return {"results": chunks}
+
+async def chat_endpoint(request):
+    data = await request.json()
+    prompt = data.get("prompt", "")
+
+    if not prompt:
+        return JSONResponse({"error": "Prompt is required"}, status_code=400)
+
+    response = await async_chat(async_client, prompt)
+    return JSONResponse({"response": response})
 
 app = Starlette(debug=True, routes=[
-    Route("/upload",      upload,       methods=["POST"]),
-    Route("/upload_path", upload_path,  methods=["POST"]),
-    Route("/search",      search,       methods=["POST"]),
+    Route("/upload",      upload,           methods=["POST"]),
+    Route("/upload_path", upload_path,      methods=["POST"]),
+    Route("/search",      search,           methods=["POST"]),
+    Route("/chat",        chat_endpoint,    methods=["POST"]),
 ])
 
 if __name__ == "__main__":

uv.lock

@@ -9,7 +9,7 @@ resolution-markers = [
 
 [[package]]
 name = "agentd"
-version = "0.1.7"
+version = "0.1.8"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "litellm" },
@@ -18,9 +18,9 @@ dependencies = [
     { name = "openai-agents" },
     { name = "pyyaml" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/1c/f6/6fc7c171e0c874e816ec063bf3cbabda4f05c6f4f1f15b65955255951027/agentd-0.1.7.tar.gz", hash = "sha256:cdcbf56dfdc0ca56067f354d890841bd71ee52210c12e468e711554007a6724c", size = 114216, upload-time = "2025-06-05T04:50:52.693Z" }
+sdist = { url = "https://files.pythonhosted.org/packages/ea/e1/a3d0d2ddb6639be34d906f13a2edc91dbf174f8dcf97a68705f3a613ff8d/agentd-0.1.8.tar.gz", hash = "sha256:9278916d228d23d67283aed0e420d14f3b6862499df5cc5a8adb92ab3583ed17", size = 114252, upload-time = "2025-07-11T16:06:57.478Z" }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/ab/c7/e786759ebd86afba150ccc1e8189e8515c760445f04374e2c4d72701c2aa/agentd-0.1.7-py3-none-any.whl", hash = "sha256:39332b8cf57dfc14b48628dfeb588afb6dbe05405945389fbe2925f3358b7c65", size = 13446, upload-time = "2025-06-05T04:50:51.046Z" },
+    { url = "https://files.pythonhosted.org/packages/d5/ca/9caa1253ab3ba151e725ea02aba334c29d659568b5341e5d886dbb394d85/agentd-0.1.8-py3-none-any.whl", hash = "sha256:15cc05ccbedfa9df8983a7a67c274c0c5a7ef029e55e6c0d7639106022c5cf06", size = 13472, upload-time = "2025-07-11T16:06:56.014Z" },
 ]
 
 [[package]]
@@ -434,7 +434,7 @@ dependencies = [
 
 [package.metadata]
 requires-dist = [
-    { name = "agentd", specifier = ">=0.1.7" },
+    { name = "agentd", specifier = ">=0.1.8" },
     { name = "aiofiles", specifier = ">=24.1.0" },
     { name = "docling", specifier = ">=2.41.0" },
     { name = "opensearch-py", extras = ["async"], specifier = ">=3.0.0" },