275 pull ollama models gui

src/client/app/content/config/tabs/models.py

@@ -13,7 +13,7 @@
 from streamlit import session_state as state
 
 from client.app.core import helpers
-from client.app.core.api import api_delete, api_get, api_post, api_put
+from client.app.core.api import api_delete, api_get, api_post, api_post_stream, api_put
 
 LOGGER = logging.getLogger("client.content.config.tabs.models")
 
@@ -324,20 +324,52 @@ def edit_model(
         st.rerun()
 
 
+#####################################################
+# Pull Dialog
+#####################################################
+@st.dialog("Pull Ollama Model")
+def pull_model_dialog(provider: str, model_id: str) -> None:
+    """Stream Ollama model pull progress."""
+    st.write(f"Pulling **{provider}/{model_id}** from Ollama registry...")
+    quoted_id = urllib.parse.quote(model_id, safe="")
+    status_text = st.empty()
+    progress_bar = st.empty()
+
+    try:
+        for event in api_post_stream(f"models/pull/{provider}/{quoted_id}"):
+            if "error" in event:
+                st.error(f"Pull failed: {event['error']}")
+                return
+            status = event.get("status", "")
+            completed = event.get("completed", 0)
+            total = event.get("total", 0)
+            if total > 0:
+                progress_bar.progress(completed / total, text=status)
+            else:
+                status_text.text(status)
+    except httpx.HTTPStatusError as exc:
+        st.error(f"Pull failed: {helpers.extract_error_detail(exc)}")
+        return
+
+    helpers.refresh_settings()
+    st.success(f"Model **{model_id}** pulled successfully. You can now enable it.")
+
+
 #####################################################
 # Table Display
 #####################################################
 def render_model_rows(model_type: str) -> None:
     """Render rows of the models."""
     models = [m for m in state["settings"]["model_configs"] if m.get("type") == model_type]
-    data_col_widths = [0.06, 0.44, 0.38, 0.12]
+    data_col_widths = [0.06, 0.40, 0.34, 0.10, 0.10]
 
     table_col_format = st.columns(data_col_widths, vertical_alignment="center")
-    col1, col2, col3, col4 = table_col_format
+    col1, col2, col3, col4, col5 = table_col_format
     col1.markdown("​", unsafe_allow_html=True, width="content")
     col2.markdown("**<u>Model</u>**", unsafe_allow_html=True)
     col3.markdown("**<u>Provider URL</u>**", unsafe_allow_html=True)
     col4.markdown("&#x200B;")
+    col5.markdown("&#x200B;")
 
     for model in models:
         model_id = model["id"]
@@ -375,6 +407,13 @@ def render_model_rows(model_type: str) -> None:
                 "model_provider": model_provider,
             },
         )
+        if model_provider == "ollama" and not model.get("usable", False):
+            col5.button(
+                "Pull",
+                on_click=pull_model_dialog,
+                key=f"runtime_{model_type}_{model_provider}_{model_id}_pull",
+                kwargs={"provider": model_provider, "model_id": model_id},
+            )
 
     if st.button(label="Add", type="primary", key=f"add_{model_type}_model"):
         edit_model(model_type=model_type, action="add")

src/client/app/core/api.py

@@ -5,12 +5,14 @@
 # spell-checker:ignore apiserver pypath
 
 import atexit
+import json
 import logging
 import os
 import secrets
 import subprocess
 import sys
 import time
+from collections.abc import Generator
 from io import TextIOWrapper
 from pathlib import Path
 from typing import Any
@@ -206,6 +208,29 @@ def api_post(
         return resp.json()
 
 
+def api_post_stream(
+    path: str,
+    json_body: dict | None = None,
+    timeout: int = 600,
+    api_prefix: str = "/v1",
+) -> Generator[dict, None, None]:
+    """Streaming POST request to the API server. Yields parsed NDJSON dicts."""
+    url = f"{_base_url(api_prefix)}/{path.lstrip('/')}"
+    with (
+        httpx.Client(headers=_headers(), timeout=timeout) as client,
+        client.stream("POST", url, json=json_body) as resp,
+    ):
+        resp.raise_for_status()
+        for raw_line in resp.iter_lines():
+            stripped = raw_line.strip()
+            if not stripped:
+                continue
+            try:
+                yield json.loads(stripped)
+            except json.JSONDecodeError:
+                continue
+
+
 def api_put(
     path: str,
     json: dict | None = None,

src/client/tests/content/config/tabs/test_models.py

@@ -998,7 +998,7 @@ def test_renders_rows_only_for_matching_type(self, make_model_state, mock_st):
         ]
         state = make_model_state(model_configs=configs)
 
-        cols = [MagicMock(), MagicMock(), MagicMock(), MagicMock()]
+        cols = [MagicMock(), MagicMock(), MagicMock(), MagicMock(), MagicMock()]
         mock_st.columns.side_effect = lambda widths, **kw: cols
         mock_st.button.return_value = False
 
@@ -1020,7 +1020,7 @@ def test_edit_button_kwargs(self, make_model_state, mock_st):
         ]
         state = make_model_state(model_configs=configs)
 
-        cols = [MagicMock(), MagicMock(), MagicMock(), MagicMock()]
+        cols = [MagicMock(), MagicMock(), MagicMock(), MagicMock(), MagicMock()]
         mock_st.columns.side_effect = lambda widths, **kw: cols
         mock_st.button.return_value = False
 

src/server/app/api/v1/endpoints/models.py

@@ -6,15 +6,18 @@
 """
 # spell-checker:ignore litellm ollama rerank
 
+import json
 from typing import Optional
 
 import litellm
 from fastapi import APIRouter, HTTPException, Query
+from fastapi.responses import StreamingResponse
 
 from server.app.core.settings import _settings_lock, settings
 from server.app.database.settings import persist_settings
 from server.app.models.connectivity import check_single_model
 from server.app.models.litellm_utils import find_model
+from server.app.models.ollama import pull_ollama_model
 from server.app.models.schemas import ModelConfig, ModelSensitive, ModelUpdate, SupportedProviderIds
 
 litellm.suppress_debug_info = True
@@ -100,6 +103,38 @@ async def models_supported(
     return _get_supported(model_provider=model_provider, model_type=model_type)
 
 
+# --- Pull endpoint (must be before /{provider}/{model_id:path}) ---
+
+
+@auth.post("/pull/{provider}/{model_id:path}")
+async def pull_model(provider: str, model_id: str):
+    """Pull an Ollama model and stream progress as NDJSON."""
+    if provider.casefold() != "ollama":
+        raise HTTPException(status_code=400, detail="Pull is only supported for Ollama models")
+
+    cfg = _find_model(provider, model_id)
+    if cfg is None:
+        raise HTTPException(status_code=404, detail=f"Model config not found: {provider}/{model_id}")
+    if not cfg.api_base:
+        raise HTTPException(status_code=400, detail=f"Model {provider}/{model_id} has no API base URL configured")
+    api_base: str = cfg.api_base
+
+    async def _stream():
+        error_occurred = False
+        async for event in pull_ollama_model(api_base, model_id):
+            if "error" in event:
+                error_occurred = True
+            yield json.dumps(event) + "\n"
+        if not error_occurred:
+            await check_single_model(cfg)
+            if not await persist_settings():
+                yield json.dumps({"error": _PERSIST_FAIL}) + "\n"
+            else:
+                yield json.dumps({"status": "success"}) + "\n"
+
+    return StreamingResponse(_stream(), media_type="application/x-ndjson")
+
+
 @auth.get("/{provider}/{model_id:path}", response_model=ModelConfig, response_model_exclude_unset=True)
 async def get_model(provider: str, model_id: str, include_sensitive: bool = Query(default=False)):
     """Return a single model configuration by provider and id (case-insensitive)."""

src/server/app/models/ollama.py

@@ -6,8 +6,10 @@
 """
 # spell-checker: ignore ollama nomic
 
+import json
 import logging
 import os
+from collections.abc import AsyncGenerator
 
 import httpx
 
@@ -129,3 +131,29 @@ async def load_ollama_models() -> None:
         LOGGER.info("Removed %d Ollama model(s) no longer pulled", len(removed))
 
     LOGGER.info("Discovered %d Ollama model(s) at %s", len(models), api_base)
+
+
+async def pull_ollama_model(api_base: str, model_name: str) -> AsyncGenerator[dict, None]:
+    """Stream pull progress from an Ollama server as dicts.
+
+    Each yielded dict mirrors the NDJSON lines from Ollama's ``/api/pull``
+    endpoint (keys like ``status``, ``completed``, ``total``, ``digest``).
+    On error an ``{"error": "..."}`` dict is yielded.
+    """
+    url = f"{api_base.rstrip('/')}/api/pull"
+    pull_read_timeout = 120.0  # Longer timeout for pull — gaps between progress lines during layer downloads
+    try:
+        async with httpx.AsyncClient(timeout=httpx.Timeout(pull_read_timeout, connect=CONNECT_TIMEOUT)) as client:  # noqa: SIM117
+            async with client.stream("POST", url, json={"name": model_name}) as resp:
+                resp.raise_for_status()
+                async for raw_line in resp.aiter_lines():
+                    stripped = raw_line.strip()
+                    if not stripped:
+                        continue
+                    try:
+                        yield json.loads(stripped)
+                    except json.JSONDecodeError:
+                        continue
+    except httpx.HTTPError as exc:
+        LOGGER.warning("Ollama pull failed for '%s' at %s: %s", model_name, api_base, exc)
+        yield {"error": str(exc)}