Start the Pydantic AI Bot on an Ollama model named with its tag (#610)

The Pydantic AI Bot builds its model id as `provider:model`, and took `BOT_MODEL` as already
carrying a provider whenever it held a colon. Ollama puts a colon in every model's name: `ollama
list` shows `llama3.1:8b`, `qwen3:8b`, `mistral:7b`, and that is the name the setup screen passes
on for an OpenAI-compatible endpoint. So the harness handed `llama3.1:8b` to Pydantic AI verbatim,
Pydantic AI read `llama3.1` as the provider and refused it, and the module failed at import with
`UserError: Unknown model: llama3.1:8b`: the container never served a run. `mistral:7b` is worse
by one step, being taken as the Mistral provider and failing for want of its package.

A model's name now carries a provider only when it begins with the chosen provider's own name and a
colon, so `anthropic:claude-sonnet-4-5` still reaches Anthropic as written and anything else is
prefixed with the provider. Pydantic AI splits on the first colon, so `openai:llama3.1:8b` reaches
the endpoint as `llama3.1:8b`.

The test follows the LlamaIndex Bot's, with the Responses API faked because that is the route
`openai:` takes in Pydantic AI. Before this change the two tagged models fail and the other four
choices pass.

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

Author: kevin9327

.github/workflows/ci.yml

@@ -304,6 +304,13 @@ jobs:
           . .venv-agno/bin/activate
           python -m pip install --requirement agent-agno/requirements.txt --requirement agent-agno/requirements-test.txt
           python -m pytest agent-agno/tests -q
+      - name: Pydantic AI model-choice regression
+        run: |
+          set -euo pipefail
+          python -m venv .venv-pydantic-ai
+          . .venv-pydantic-ai/bin/activate
+          python -m pip install --requirement agent-pydantic-ai/requirements.txt --requirement agent-pydantic-ai/requirements-test.txt
+          python -m pytest agent-pydantic-ai/tests -q
       - run: bun install --frozen-lockfile
       - run: bun test tests/compose.test.ts
       - run: docker compose --env-file /dev/null --profile harness config --format json >/dev/null

CHANGELOG.md

@@ -16,6 +16,14 @@ Compose passes when the setup screen names none, so the run ended in `Unsupporte
 parameter the model does not take is now dropped instead, the way the LlamaIndex Bot already does
 it. The Anthropic key and an OpenAI-compatible endpoint behave as before.
 
+### The Pydantic AI Bot starts on an Ollama model named with its tag
+
+Ollama names every model with a tag after a colon, as in `llama3.1:8b`, and that is the name the
+setup screen passes on for an OpenAI-compatible endpoint. The Pydantic AI Bot took any colon in the
+model's name to mean the name already carried a provider, so Pydantic AI read `llama3.1` as one,
+refused it as unknown, and the Bot never started. A model's name now carries a provider only when it
+begins with the chosen provider's own, as in `anthropic:claude-sonnet-4-5`.
+
 ## 0.0.13
 
 ### Fresh desktop setup installs its runtime before sign-in

agent-pydantic-ai/requirements-test.txt

@@ -0,0 +1,2 @@
+httpx==0.28.1
+pytest==9.0.2

agent-pydantic-ai/src/main.py

@@ -15,10 +15,15 @@
 
 
 def _model_id() -> str:
-    """`provider:model`, which is the form Pydantic AI names a model in."""
+    """`provider:model`, which is the form Pydantic AI names a model in.
+
+    A colon in `BOT_MODEL` names the provider only when it follows the provider's own name. Any
+    other colon is part of the model's name: Ollama tags every model with one, as in `llama3.1:8b`,
+    and Pydantic AI read the part before it as a provider, refused an unknown one and started no Bot.
+    """
     provider = (os.environ.get("BOT_PROVIDER") or "openai").strip()
     model = (os.environ.get("BOT_MODEL") or "gpt-4o-mini").strip()
-    return model if ":" in model else f"{provider}:{model}"
+    return model if model.startswith(f"{provider}:") else f"{provider}:{model}"
 
 
 agent = Agent(_model_id())

agent-pydantic-ai/tests/test_main.py

@@ -0,0 +1,203 @@
+import importlib
+import json
+import socket
+import sys
+import threading
+import time
+from pathlib import Path
+
+import pytest
+import uvicorn
+from fastapi import FastAPI, Request
+from fastapi.responses import StreamingResponse
+from fastapi.testclient import TestClient
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
+
+TOKEN = "test-token"
+RUN = {
+    "threadId": "thread-1",
+    "runId": "run-1",
+    "state": {},
+    "messages": [{"id": "m1", "role": "user", "content": "Say hello"}],
+    "tools": [],
+    "context": [],
+    "forwardedProps": {},
+}
+
+
+def _sse(events):
+    async def stream():
+        for name, data in events:
+            yield f"event: {name}\ndata: {json.dumps(data)}\n\n"
+
+    return StreamingResponse(stream(), media_type="text/event-stream")
+
+
+def _provider_app(seen):
+    app = FastAPI()
+
+    # `openai:` is the Responses API in Pydantic AI, so that is the route an OpenAI key and an
+    # OpenAI-compatible endpoint both reach.
+    @app.post("/v1/responses")
+    async def openai_responses(request: Request):
+        body = await request.json()
+        seen.append(("openai", body["model"]))
+        response = {
+            "id": "resp",
+            "object": "response",
+            "created_at": 0,
+            "model": body["model"],
+            "status": "in_progress",
+            "output": [],
+            "parallel_tool_calls": True,
+            "tool_choice": "auto",
+            "tools": [],
+        }
+        item = {"id": "msg", "type": "message", "role": "assistant", "status": "completed"}
+        text = {"type": "output_text", "text": "hello", "annotations": []}
+        done = {
+            **response,
+            "status": "completed",
+            "output": [{**item, "content": [text]}],
+            "usage": {
+                "input_tokens": 1,
+                "input_tokens_details": {"cached_tokens": 0},
+                "output_tokens": 1,
+                "output_tokens_details": {"reasoning_tokens": 0},
+                "total_tokens": 2,
+            },
+        }
+        return _sse(
+            [
+                ("response.created", {"type": "response.created", "sequence_number": 0, "response": response}),
+                ("response.output_item.added", {"type": "response.output_item.added", "sequence_number": 1, "output_index": 0, "item": {**item, "status": "in_progress", "content": []}}),
+                ("response.output_text.delta", {"type": "response.output_text.delta", "sequence_number": 2, "item_id": "msg", "output_index": 0, "content_index": 0, "delta": "hello", "logprobs": []}),
+                ("response.output_item.done", {"type": "response.output_item.done", "sequence_number": 3, "output_index": 0, "item": {**item, "content": [text]}}),
+                ("response.completed", {"type": "response.completed", "sequence_number": 4, "response": done}),
+            ]
+        )
+
+    @app.post("/v1/messages")
+    async def anthropic_messages(request: Request):
+        body = await request.json()
+        seen.append(("anthropic", body["model"]))
+        message = {
+            "id": "msg",
+            "type": "message",
+            "role": "assistant",
+            "model": body["model"],
+            "stop_sequence": None,
+        }
+        return _sse(
+            [
+                ("message_start", {"type": "message_start", "message": {**message, "content": [], "stop_reason": None, "usage": {"input_tokens": 1, "output_tokens": 0}}}),
+                ("content_block_start", {"type": "content_block_start", "index": 0, "content_block": {"type": "text", "text": ""}}),
+                ("content_block_delta", {"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": "hello"}}),
+                ("content_block_stop", {"type": "content_block_stop", "index": 0}),
+                ("message_delta", {"type": "message_delta", "delta": {"stop_reason": "end_turn", "stop_sequence": None}, "usage": {"output_tokens": 1}}),
+                ("message_stop", {"type": "message_stop"}),
+            ]
+        )
+
+    return app
+
+
+@pytest.fixture
+def provider():
+    seen = []
+    with socket.socket() as probe:
+        probe.bind(("127.0.0.1", 0))
+        port = probe.getsockname()[1]
+    server = uvicorn.Server(
+        uvicorn.Config(_provider_app(seen), host="127.0.0.1", port=port, log_level="error")
+    )
+    thread = threading.Thread(target=server.run, daemon=True)
+    thread.start()
+    deadline = time.monotonic() + 10
+    while not server.started and time.monotonic() < deadline:
+        time.sleep(0.01)
+    yield f"http://127.0.0.1:{port}", seen
+    server.should_exit = True
+    thread.join(timeout=10)
+
+
+def _endpoint(model):
+    """What the desktop writes for an OpenAI-compatible endpoint serving `model`."""
+    return (
+        lambda base: {
+            "BOT_PROVIDER": "",
+            "BOT_MODEL": model,
+            "OPENAI_API_KEY": "no-key-needed",
+            "OPENAI_BASE_URL": f"{base}/v1",
+            "ANTHROPIC_API_KEY": "",
+        },
+        ("openai", model),
+    )
+
+
+CHOICES = {
+    "an Anthropic key": (
+        lambda base: {
+            "BOT_PROVIDER": "anthropic",
+            "BOT_MODEL": "claude-sonnet-4-5",
+            "ANTHROPIC_API_KEY": "test-key",
+            "ANTHROPIC_BASE_URL": base,
+            "OPENAI_API_KEY": "",
+            "OPENAI_BASE_URL": "",
+        },
+        ("anthropic", "claude-sonnet-4-5"),
+    ),
+    # Written the way Pydantic AI names a model, which still reaches that provider.
+    "a model already named with its provider": (
+        lambda base: {
+            "BOT_PROVIDER": "anthropic",
+            "BOT_MODEL": "anthropic:claude-sonnet-4-5",
+            "ANTHROPIC_API_KEY": "test-key",
+            "ANTHROPIC_BASE_URL": base,
+            "OPENAI_API_KEY": "",
+            "OPENAI_BASE_URL": "",
+        },
+        ("anthropic", "claude-sonnet-4-5"),
+    ),
+    "an OpenAI-compatible endpoint": _endpoint("local-model"),
+    # Ollama names every model with its tag after a colon, which is the separator Pydantic AI puts
+    # between a provider and a model.
+    "an Ollama model named with its tag": _endpoint("llama3.1:8b"),
+    # And a tag on a model that shares its name with a provider Pydantic AI knows.
+    "an Ollama model named like a provider": _endpoint("mistral:7b"),
+    "an OpenAI key": (
+        lambda base: {
+            "BOT_PROVIDER": "",
+            "BOT_MODEL": "gpt-5.5",
+            "OPENAI_API_KEY": "test-key",
+            "OPENAI_BASE_URL": f"{base}/v1",
+            "ANTHROPIC_API_KEY": "",
+        },
+        ("openai", "gpt-5.5"),
+    ),
+}
+
+
+@pytest.mark.parametrize("choice", list(CHOICES))
+def test_a_run_reaches_the_model_the_setup_screen_chose(monkeypatch, provider, choice):
+    base, seen = provider
+    environment, expected = CHOICES[choice]
+    monkeypatch.delenv("OPENAI_API_BASE", raising=False)
+    monkeypatch.setenv("MANAGED_AGENT_TOKEN", TOKEN)
+    monkeypatch.setenv("PYDANTIC_AI_NO_BANNER", "1")
+    for key, value in environment(base).items():
+        monkeypatch.setenv(key, value)
+
+    from src import main
+
+    main = importlib.reload(main)
+    response = TestClient(main.app).post(
+        "/", json=RUN, headers={"x-openbot-agent-token": TOKEN}
+    )
+
+    assert response.status_code == 200
+    assert '"RUN_FINISHED"' in response.text
+    assert '"RUN_ERROR"' not in response.text
+    assert "hello" in response.text
+    assert seen == [expected]