feat: add provider-neutral LLM factory with an OpenAI-compatible provider (fc5fdd39) · Commits · OOP / code / ai2

ai_agent/env.py

0 → 100644

+8 −0

Changes for ai_agent/env.py: 8 added lines, 0 removed lines.
Original line number	Diff line number	Diff line
		import os


		def require_env(name: str) -> str:
		value = os.getenv(name)
		if value is None or not value.strip():
		raise OSError(f"{name} environment variable is not set")
		return value.strip()

ai_agent/llm.py

0 → 100644

+59 −0

Changes for ai_agent/llm.py: 59 added lines, 0 removed lines.
Original line number	Diff line number	Diff line
		import os

		from dotenv import load_dotenv
		from langchain_core.language_models.chat_models import BaseChatModel
		from langchain_groq import ChatGroq
		from langchain_openai import ChatOpenAI
		from pydantic import SecretStr

		from ai_agent.env import require_env

		load_dotenv()

		DEFAULT_PROVIDER = "groq"
		DEFAULT_GROQ_MODEL = "openai/gpt-oss-20b"
		DEFAULT_OPENAI_TIMEOUT = 120.0


		def _optional_env(name: str, default: str) -> str:
		return (os.getenv(name) or "").strip() or default


		def _create_groq_llm() -> BaseChatModel:
		return ChatGroq(
		model=_optional_env("GROQ_MODEL_NAME", DEFAULT_GROQ_MODEL),
		temperature=0,
		max_tokens=None,
		reasoning_format="parsed",
		timeout=15,
		max_retries=3,
		api_key=SecretStr(require_env("GROQ_API_KEY")),
		)


		def _create_openai_llm() -> BaseChatModel:
		"""Any OpenAI-compatible endpoint, e.g. Open WebUI (https://host/api) or vLLM."""
		raw_timeout = _optional_env("OPENAI_TIMEOUT", str(DEFAULT_OPENAI_TIMEOUT))
		try:
		timeout = float(raw_timeout)
		except ValueError as e:
		raise OSError(f"OPENAI_TIMEOUT must be a number, got {raw_timeout!r}") from e

		return ChatOpenAI(
		model=require_env("OPENAI_MODEL_NAME"),
		base_url=require_env("OPENAI_BASE_URL"),
		api_key=SecretStr(require_env("OPENAI_API_KEY")),
		temperature=0,
		timeout=timeout,
		max_retries=2,
		)


		def create_llm() -> BaseChatModel:
		"""Create the chat model selected by LLM_PROVIDER (groq by default)."""
		provider = _optional_env("LLM_PROVIDER", DEFAULT_PROVIDER).lower()
		if provider == "groq":
		return _create_groq_llm()
		if provider == "openai":
		return _create_openai_llm()
		raise OSError(f"LLM_PROVIDER must be 'groq' or 'openai', got {provider!r}")

ai_agent/requirements.txt

+1 −0

Changes for ai_agent/requirements.txt: 1 added line, 0 removed lines.
Original line number	Diff line number	Diff line
		@@ -6,3 +6,4 @@ python-dotenv==1.2.1
		fastapi==0.129.0
		uvicorn==0.40.0
		redis==5.2.1
		langchain-openai==1.7.3

ai_agent/routes/groq.py

+3 −3

Changes for ai_agent/routes/groq.py: 3 added lines, 3 removed lines.
Original line number	Diff line number	Diff line
		@@ -6,7 +6,7 @@ from fastapi.responses import JSONResponse, StreamingResponse

		from ai_agent.memory import RedisShortTermMemory
		from ai_agent.security import get_authorization
		from ai_agent.sub_agents.groq_agent import create_groq_agent
		from ai_agent.sub_agents.agent import create_agent
		from ai_agent.utils import stream_agent_response

		logger = logging.getLogger(__name__)
		@@ -26,7 +26,7 @@ async def groq_query(
		if not query:
		return JSONResponse(status_code=400, content={"error": "No query provided"})

		agent = await create_groq_agent(authorization)
		agent = await create_agent(authorization)
		try:
		effective_query = await memory_store.augment_query(query, session_id)
		result = await agent.run(effective_query)
		@@ -51,7 +51,7 @@ async def groq_stream(
		if not query:
		return JSONResponse(status_code=400, content={"error": "No query provided"})

		agent = await create_groq_agent(authorization)
		agent = await create_agent(authorization)

		try:
		effective_query = await memory_store.augment_query(query, session_id)

ai_agent/sub_agents/agent.py

0 → 100644

+48 −0

Changes for ai_agent/sub_agents/agent.py: 48 added lines, 0 removed lines.
Original line number	Diff line number	Diff line
		import logging

		from dotenv import load_dotenv
		from mcp_use import MCPAgent, MCPClient # type: ignore[import-untyped]

		from ai_agent.env import require_env
		from ai_agent.llm import create_llm
		from ai_agent.prompts import CONVERSATION_MEMORY_PROMPT

		load_dotenv()
		logging.getLogger("mcp_use").setLevel(logging.ERROR)

		logger = logging.getLogger(__name__)


		async def create_agent(authorization: str \| None):
		"""Create an MCPAgent backed by the configured LLM provider."""
		try:
		mcp_server_url = require_env("MCP_SERVER_URL")
		llm = create_llm()
		except OSError as e:
		logger.error("Agent configuration error: %s", e)
		raise

		client = MCPClient(
		{
		"mcpServers": {
		"http": {
		"url": mcp_server_url,
		"headers": (
		{"Authorization": authorization} if authorization else {}
		),
		# mcp_use's default ({}) runs OAuth discovery on every connect.
		"auth": None,
		}
		},
		}
		)
		await client.create_session("http")

		agent = MCPAgent(
		client=client,
		max_steps=20,
		system_prompt=CONVERSATION_MEMORY_PROMPT,
		llm=llm,
		verbose=True,
		)
		return agent