Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

feat(embeddings): make Ollama embed timeout configurable via env (#656)

Austin Travereef02da55ffe
diff --git a/docs/CONFIG.md b/docs/CONFIG.md
index 57b2254af..ed4821615 100644
--- a/docs/CONFIG.md
+++ b/docs/CONFIG.md
@@ -229,6 +229,7 @@ The embedder is separate from the LLM provider.
|----------|---------|-------|
| `gemini` | `GEMINI_API_KEY` | Default when key is present |
| `openai` | `OPENAI_API_KEY` | OpenAI `text-embedding-3-small` |
+| `ollama` | `OLLAMA_EMBEDDING_MODEL` | Local Ollama embeddings, no API key |
| `mock` | n/a | Dummy embeddings, no semantic search |
```bash
@@ -258,6 +259,9 @@ The `.repowise/.env` file is gitignored automatically.
| `OPENAI_API_KEY` | OpenAI API key |
| `GEMINI_API_KEY` / `GOOGLE_API_KEY` | Google Gemini API key |
| `OLLAMA_BASE_URL` | Ollama server URL (default: `http://localhost:11434`) |
+| `OLLAMA_EMBEDDING_MODEL` | Ollama embedding model (also selects the `ollama` embedder) |
+| `OLLAMA_EMBEDDING_DIMS` | Ollama embedding output dimensions (optional; inferred from the model otherwise) |
+| `OLLAMA_EMBEDDING_TIMEOUT` | Ollama embed request timeout in seconds (default: `30`); raise it for long pages on slow local models |
| `LITELLM_API_KEY` | LiteLLM proxy key |
| `REPOWISE_PROVIDER` | Override provider (skips auto-detection) |
| `REPOWISE_EMBEDDER` | Embedder: `gemini`, `openai`, or `mock` |
diff --git a/packages/cli/src/repowise/cli/providers/embedders.py b/packages/cli/src/repowise/cli/providers/embedders.py
index ee51bbef6..17bf2a451 100644
--- a/packages/cli/src/repowise/cli/providers/embedders.py
+++ b/packages/cli/src/repowise/cli/providers/embedders.py
@@ -15,10 +15,15 @@ def _embedder_kwargs(embedder_name: str) -> dict[str, Any]:
dimensions = os.environ.get("OLLAMA_EMBEDDING_DIMS") or os.environ.get(
"REPOWISE_EMBEDDING_DIMS"
)
+ timeout = os.environ.get("OLLAMA_EMBEDDING_TIMEOUT") or os.environ.get(
+ "REPOWISE_EMBEDDING_TIMEOUT"
+ )
if base_url:
kwargs["base_url"] = base_url
if dimensions:
kwargs["dimensions"] = int(dimensions)
+ if timeout:
+ kwargs["timeout"] = float(timeout)
elif embedder_name == "gemini":
dimensions = os.environ.get("REPOWISE_EMBEDDING_DIMS")
if dimensions:
diff --git a/packages/core/src/repowise/core/providers/embedding/ollama.py b/packages/core/src/repowise/core/providers/embedding/ollama.py
index 8c188ffdd..bf09ceabd 100644
--- a/packages/core/src/repowise/core/providers/embedding/ollama.py
+++ b/packages/core/src/repowise/core/providers/embedding/ollama.py
@@ -47,6 +47,10 @@ class OllamaEmbedder:
base_url: Ollama server URL. Defaults to ``http://localhost:11434``.
dimensions: Optional output dimension hint. Also sent to Ollama as
``dimensions`` when provided.
+ timeout: Per-request timeout in seconds. Falls back to the
+ ``OLLAMA_EMBEDDING_TIMEOUT`` / ``REPOWISE_EMBEDDING_TIMEOUT`` env
+ vars, then ``30.0``. Raise it when embedding long pages on a slow
+ local model that would otherwise exceed the default and be dropped.
"""
def __init__(
@@ -54,7 +58,7 @@ class OllamaEmbedder:
model: str | None = None,
base_url: str | None = None,
dimensions: int | None = None,
- timeout: float = _DEFAULT_TIMEOUT,
+ timeout: float | None = None,
) -> None:
self._model = (
model
@@ -70,7 +74,14 @@ class OllamaEmbedder:
)
self._requested_dimensions = dimensions or (int(env_dimensions) if env_dimensions else None)
self._dimensions = self._requested_dimensions or _infer_dimensions(self._model)
- self._timeout = timeout
+ env_timeout = os.environ.get("OLLAMA_EMBEDDING_TIMEOUT") or os.environ.get(
+ "REPOWISE_EMBEDDING_TIMEOUT"
+ )
+ self._timeout = (
+ timeout
+ if timeout is not None
+ else (float(env_timeout) if env_timeout else _DEFAULT_TIMEOUT)
+ )
@property
def dimensions(self) -> int:
diff --git a/tests/unit/cli/test_shared_helpers.py b/tests/unit/cli/test_shared_helpers.py
index 6fca2ecc3..e6f716609 100644
--- a/tests/unit/cli/test_shared_helpers.py
+++ b/tests/unit/cli/test_shared_helpers.py
@@ -86,6 +86,16 @@ def test_build_embedder_supports_ollama(monkeypatch: pytest.MonkeyPatch) -> None
assert embedder._model == "qwen3-embedding:0.6b"
+def test_build_embedder_ollama_timeout_from_env(monkeypatch: pytest.MonkeyPatch) -> None:
+ from repowise.core.providers.embedding.ollama import OllamaEmbedder
+
+ monkeypatch.setenv("OLLAMA_EMBEDDING_MODEL", "qwen3-embedding:0.6b")
+ monkeypatch.setenv("OLLAMA_EMBEDDING_TIMEOUT", "300")
+ embedder = providers.build_embedder("ollama")
+ assert isinstance(embedder, OllamaEmbedder)
+ assert embedder._timeout == 300.0
+
+
def test_build_vector_store_returns_a_store(tmp_path) -> None:
from repowise.core.providers.embedding.base import MockEmbedder
diff --git a/tests/unit/test_persistence/test_ollama_embedder.py b/tests/unit/test_persistence/test_ollama_embedder.py
index ffc7bdfa2..f0ff13c09 100644
--- a/tests/unit/test_persistence/test_ollama_embedder.py
+++ b/tests/unit/test_persistence/test_ollama_embedder.py
@@ -118,3 +118,34 @@ def test_qwen3_dimensions_are_inferred() -> None:
assert OllamaEmbedder(model="qwen3-embedding:0.6b").dimensions == 1024
assert OllamaEmbedder(model="qwen3-embedding:4b").dimensions == 2560
assert OllamaEmbedder(model="qwen3-embedding:8b").dimensions == 4096
+
+
+def test_timeout_defaults_to_thirty_seconds(monkeypatch: pytest.MonkeyPatch) -> None:
+ monkeypatch.delenv("OLLAMA_EMBEDDING_TIMEOUT", raising=False)
+ monkeypatch.delenv("REPOWISE_EMBEDDING_TIMEOUT", raising=False)
+ assert OllamaEmbedder(model="embeddinggemma")._timeout == 30.0
+
+
+def test_timeout_can_come_from_env(monkeypatch: pytest.MonkeyPatch) -> None:
+ monkeypatch.setenv("OLLAMA_EMBEDDING_TIMEOUT", "120")
+ assert OllamaEmbedder(model="embeddinggemma")._timeout == 120.0
+
+
+def test_repowise_embedding_timeout_is_a_fallback(monkeypatch: pytest.MonkeyPatch) -> None:
+ monkeypatch.delenv("OLLAMA_EMBEDDING_TIMEOUT", raising=False)
+ monkeypatch.setenv("REPOWISE_EMBEDDING_TIMEOUT", "90")
+ assert OllamaEmbedder(model="embeddinggemma")._timeout == 90.0
+
+
+def test_explicit_timeout_overrides_env(monkeypatch: pytest.MonkeyPatch) -> None:
+ monkeypatch.setenv("OLLAMA_EMBEDDING_TIMEOUT", "120")
+ assert OllamaEmbedder(model="embeddinggemma", timeout=5.0)._timeout == 5.0
+
+
+async def test_env_timeout_is_applied_to_the_request(monkeypatch: pytest.MonkeyPatch) -> None:
+ monkeypatch.setenv("OLLAMA_EMBEDDING_TIMEOUT", "300")
+ embedder = OllamaEmbedder(model="embeddinggemma")
+
+ await embedder.embed(["one"])
+
+ assert _FakeAsyncClient.calls[0]["timeout"] == 300.0