feat(component): add Ollama LLM and Embedding providers

src/everos/component/embedding/__init__.py

@@ -8,6 +8,8 @@
 - :class:`EmbeddingError` — backward-compat alias for ``EmbeddingServiceError``.
 - :class:`OpenAIEmbeddingProvider` — concrete provider for any
   OpenAI-protocol embeddings endpoint (DeepInfra, vLLM, OpenAI, …).
+- :class:`OllamaEmbeddingProvider` — local Ollama embedding provider
+  with sensible defaults (``http://localhost:11434/v1``, dim=768).
 - :func:`build_embedding_provider` — settings-driven factory.
 
 External usage::
@@ -22,6 +24,7 @@
 from .accessor import EmbeddingNotConfiguredError as EmbeddingNotConfiguredError
 from .accessor import get_embedder as get_embedder
 from .factory import build_embedding_provider as build_embedding_provider
+from .ollama_provider import OllamaEmbeddingProvider as OllamaEmbeddingProvider
 from .openai_provider import OpenAIEmbeddingProvider as OpenAIEmbeddingProvider
 from .protocol import EmbeddingError as EmbeddingError
 from .protocol import EmbeddingProvider as EmbeddingProvider
@@ -31,6 +34,7 @@
     "EmbeddingNotConfiguredError",
     "EmbeddingProvider",
     "EmbeddingServiceError",
+    "OllamaEmbeddingProvider",
     "OpenAIEmbeddingProvider",
     "build_embedding_provider",
     "get_embedder",

src/everos/component/embedding/ollama_provider.py

@@ -0,0 +1,66 @@
+"""Ollama embedding provider.
+
+Ollama exposes an OpenAI-compatible ``/v1/embeddings`` endpoint at
+``http://localhost:11434/v1``. This provider applies Ollama-specific
+defaults so callers get a working configuration out of the box — no
+need to remember the base URL or supply a dummy API key.
+
+Usage::
+    from everos.component.embedding.ollama_provider import OllamaEmbeddingProvider
+
+    provider = OllamaEmbeddingProvider(
+        model="nomic-embed-text",
+        dim=768,
+    )
+    vec = await provider.embed("hello world")
+"""
+
+from __future__ import annotations
+
+from .openai_provider import OpenAIEmbeddingProvider
+
+
+_OLLAMA_DEFAULT_BASE_URL = "http://localhost:11434/v1"
+_OLLAMA_PLACEHOLDER_KEY = "ollama"
+
+
+class OllamaEmbeddingProvider(OpenAIEmbeddingProvider):
+    """Ollama embedding provider — OpenAI-compatible with local defaults.
+
+    Same semantics as :class:`OpenAIEmbeddingProvider` except
+    ``api_key`` and ``base_url`` fall back to Ollama's conventional
+    values when omitted.
+
+    Args:
+        model: Ollama model id (e.g. ``"nomic-embed-text"``).
+        api_key: API key; defaults to ``"ollama"``.
+        base_url: Endpoint; defaults to ``http://localhost:11434/v1``.
+        dim: Target vector dimension; defaults to 768 (Ollama's typical).
+        timeout: Per-request timeout, seconds.
+        max_retries: Retry budget.
+        batch_size: Embeddings per ``/embeddings`` call.
+        max_concurrent: Cap on in-flight chunked requests.
+    """
+
+    def __init__(
+        self,
+        *,
+        model: str,
+        api_key: str = _OLLAMA_PLACEHOLDER_KEY,
+        base_url: str = _OLLAMA_DEFAULT_BASE_URL,
+        dim: int = 768,
+        timeout: float = 30.0,
+        max_retries: int = 3,
+        batch_size: int = 10,
+        max_concurrent: int = 5,
+    ) -> None:
+        super().__init__(
+            model=model,
+            api_key=api_key,
+            base_url=base_url,
+            dim=dim,
+            timeout=timeout,
+            max_retries=max_retries,
+            batch_size=batch_size,
+            max_concurrent=max_concurrent,
+        )

src/everos/component/llm/__init__.py

@@ -11,6 +11,8 @@
 - :class:`LLMNotConfiguredError` — raised when credentials are missing.
 - :class:`OpenAIProvider` — concrete provider wrapping
   ``openai.AsyncOpenAI`` against any OpenAI-compatible endpoint.
+- :class:`OllamaProvider` — local Ollama LLM provider with sensible
+  defaults (``http://localhost:11434/v1``, ``"ollama"`` key).
 - :func:`build_llm_provider` — settings-driven factory.
 - :func:`get_llm_client` — process-wide lazy singleton accessor.
 
@@ -24,6 +26,7 @@
 from .client import get_llm_client as get_llm_client
 from .client import get_multimodal_llm_client as get_multimodal_llm_client
 from .factory import build_llm_provider as build_llm_provider
+from .ollama_provider import OllamaProvider as OllamaProvider
 from .openai_provider import OpenAIProvider as OpenAIProvider
 from .protocol import ChatMessage as ChatMessage
 from .protocol import ChatResponse as ChatResponse
@@ -37,6 +40,7 @@
     "LLMClient",
     "LLMError",
     "LLMNotConfiguredError",
+    "OllamaProvider",
     "OpenAIProvider",
     "Usage",
     "build_llm_provider",

src/everos/component/llm/ollama_provider.py

@@ -0,0 +1,61 @@
+"""Ollama LLM provider for everos.
+
+Ollama exposes an OpenAI-compatible API at ``http://localhost:11434/v1``
+by default, with ``api-key`` accepting any non-empty value. This
+provider extends :class:`OpenAIProvider` with Ollama-specific defaults
+so callers don't need to remember the canonical base URL or dummy key.
+
+Usage::
+    settings = LLMSettings(
+        model="llama3.1",
+        base_url="http://localhost:11434/v1",  # or omit — Ollama default
+        api_key=SecretStr("ollama"),            # or omit — Ollama default
+    )
+    provider = OllamaProvider(model=settings.model)
+"""
+
+from __future__ import annotations
+
+from typing import Any
+
+from .openai_provider import OpenAIProvider
+from .protocol import ChatMessage, ChatResponse
+
+
+_OLLAMA_DEFAULT_BASE_URL = "http://localhost:11434/v1"
+_OLLAMA_PLACEHOLDER_KEY = "ollama"
+
+
+class OllamaProvider(OpenAIProvider):
+    """Ollama LLM provider — OpenAI-compatible with local defaults.
+
+    Identical to :class:`OpenAIProvider` except that ``api_key`` and
+    ``base_url`` fall back to Ollama's conventional values when omitted.
+
+    Args:
+        model: Ollama model id (e.g. ``"llama3.1"``, ``"qwen2.5"``).
+        api_key: API key; defaults to ``"ollama"`` (Ollama's placeholder).
+        base_url: Endpoint; defaults to ``http://localhost:11434/v1``.
+        timeout: Per-request timeout in seconds.
+        temperature: Default sampling temperature.
+        max_tokens: Default max-tokens cap.
+    """
+
+    def __init__(
+        self,
+        *,
+        model: str,
+        api_key: str = _OLLAMA_PLACEHOLDER_KEY,
+        base_url: str | None = _OLLAMA_DEFAULT_BASE_URL,
+        timeout: float = 60.0,
+        temperature: float = 0.0,
+        max_tokens: int | None = None,
+    ) -> None:
+        super().__init__(
+            model=model,
+            api_key=api_key,
+            base_url=base_url,
+            timeout=timeout,
+            temperature=temperature,
+            max_tokens=max_tokens,
+        )

tests/unit/test_component/test_ollama_providers.py

@@ -0,0 +1,106 @@
+"""Tests for Ollama LLM and Embedding providers."""
+
+from __future__ import annotations
+
+import sys
+from unittest.mock import AsyncMock, patch
+
+import pytest
+
+# ── LLM Provider ────────────────────────────────────────────────────
+
+
+class TestOllamaProvider:
+    """Ollama LLM provider tests."""
+
+    def test_defaults(self):
+        """OllamaProvider applies localhost defaults when omitted."""
+        from everos.component.llm.ollama_provider import OllamaProvider
+
+        with patch("openai.AsyncOpenAI"):
+            p = OllamaProvider(model="llama3.1")
+
+        assert p._model == "llama3.1"
+
+    def test_custom_base_url(self):
+        """Custom base_url and api_key are respected."""
+        from everos.component.llm.ollama_provider import OllamaProvider
+
+        with patch("openai.AsyncOpenAI"):
+            p = OllamaProvider(
+                model="qwen2.5",
+                api_key="custom-key",
+                base_url="http://192.168.1.100:11434/v1",
+            )
+
+        assert p._model == "qwen2.5"
+
+    def test_extends_openai_provider(self):
+        """OllamaProvider inherits from OpenAIProvider."""
+        from everos.component.llm.ollama_provider import OllamaProvider
+        from everos.component.llm.openai_provider import OpenAIProvider
+
+        assert issubclass(OllamaProvider, OpenAIProvider)
+
+
+# ── Embedding Provider ──────────────────────────────────────────────
+
+
+class TestOllamaEmbeddingProvider:
+    """Ollama Embedding provider tests."""
+
+    def test_defaults(self):
+        """OllamaEmbeddingProvider applies localhost + dim=768 defaults."""
+        from everos.component.embedding.ollama_provider import (
+            OllamaEmbeddingProvider,
+        )
+
+        with patch("openai.AsyncOpenAI"):
+            p = OllamaEmbeddingProvider(model="nomic-embed-text")
+
+        assert p._model == "nomic-embed-text"
+        assert p.dim == 768
+
+    def test_custom_dim(self):
+        """Custom dim overrides the default 768."""
+        from everos.component.embedding.ollama_provider import (
+            OllamaEmbeddingProvider,
+        )
+
+        with patch("openai.AsyncOpenAI"):
+            p = OllamaEmbeddingProvider(
+                model="mxbai-embed-large",
+                dim=1024,
+            )
+
+        assert p.dim == 1024
+
+    def test_extends_openai_provider(self):
+        """OllamaEmbeddingProvider inherits from OpenAIEmbeddingProvider."""
+        from everos.component.embedding.ollama_provider import (
+            OllamaEmbeddingProvider,
+        )
+        from everos.component.embedding.openai_provider import (
+            OpenAIEmbeddingProvider,
+        )
+
+        assert issubclass(OllamaEmbeddingProvider, OpenAIEmbeddingProvider)
+
+
+# ── Import surface ──────────────────────────────────────────────────
+
+
+class TestImportSurface:
+    """Verify public imports work from the package level."""
+
+    def test_llm_ollama_importable(self):
+        """OllamaProvider is importable from everos.component.llm."""
+        from everos.component.llm import OllamaProvider
+
+        assert OllamaProvider is not None
+
+    def test_embedding_ollama_importable(self):
+        """OllamaEmbeddingProvider is importable from everos.component.embedding."""
+        from everos.component.embedding import OllamaEmbeddingProvider
+
+        assert OllamaEmbeddingProvider is not None