feat(providers): ollama cloud + python sdk

.env.template

@@ -1,3 +1,4 @@
 # LLM API Keys
 ANTHROPIC_API_KEY=sk-ant-your-key-here
 OPENAI_API_KEY=sk-your-key-here
+OLLAMA_API_KEY=you-key-here

example.py

@@ -5,6 +5,7 @@
 
 from harvestor import Harvestor, list_models
 
+
 load_dotenv()
 
 
@@ -18,24 +19,33 @@ class SimpleInvoiceSchema(BaseModel):
         None, description="The customer firstname"
     )
     customer_lastname: Optional[str] = Field(None, description="The customer lastname")
+    invoice_total_price_with_taxes: Optional[float] = Field(
+        None, description="The total price with taxes"
+    )
+    invoice_total_price_without_taxes: Optional[float] = Field(
+        None, description="The total price without taxes"
+    )
 
 
 # List available models
 print("Available models:", list(list_models().keys()))
 
 # Use default model (claude-haiku)
-h = Harvestor(model="claude-haiku")
+# h = Harvestor(model="claude-haiku")
 
-output = h.harvest_file(
-    source="data/uploads/keep_for_test.jpg", schema=SimpleInvoiceSchema
-)
+# output = h.harvest_file(
+#     source="data/uploads/keep_for_test.jpg", schema=SimpleInvoiceSchema
+# )
 
-print(output.to_summary())
+# print(output.to_summary())
 
 # Alternative: use OpenAI
 # h_openai = Harvestor(model="gpt-4o-mini")
-# output = h_openai.harvest_file("invoice.jpg", schema=SimpleInvoiceSchema)
+# output = h_openai.harvest_file("data/uploads/keep_for_test.jpg", schema=SimpleInvoiceSchema)
 
-# Alternative: use local Ollama (free)
-# h_ollama = Harvestor(model="llava")
-# output = h_ollama.harvest_file("invoice.jpg", schema=SimpleInvoiceSchema)
+# Alternative: use local Ollama (free) or cloud Ollama
+h_ollama = Harvestor(model="gemma3:4b-cloud")
+output = h_ollama.harvest_file(
+    "data/uploads/keep_for_test.jpg", schema=SimpleInvoiceSchema
+)
+print(output.to_summary())

pyproject.toml

@@ -26,7 +26,7 @@ dependencies = [
     "langchain-openai>=0.0.5",
     "anthropic>=0.18.0",
     "openai>=1.10.0",
-    "httpx>=0.27.0",  # For Ollama provider
+    "ollama>=0.6.1",
 
     # Document Processing
     "PyMuPDF>=1.23.0",

src/harvestor/providers/ollama.py

@@ -6,7 +6,7 @@
 import os
 from typing import Optional
 
-import httpx
+from ollama import generate, Client, list as list_models
 
 from .base import BaseLLMProvider, CompletionResult, ModelInfo
 
@@ -66,17 +66,25 @@ def __init__(
         if model not in OLLAMA_MODELS:
             # Allow custom models not in the predefined list
             self.model_config = {
-                "id": f"{model}:latest" if ":" not in model else model,
+                "id": f"{model}:latest"
+                if ":" not in model
+                else model,  # allow models like ministral-3:3b
                 "input_cost": 0.0,
                 "output_cost": 0.0,
-                "supports_vision": False,
+                "supports_vision": True,
                 "context_window": 8192,
             }
         else:
             self.model_config = OLLAMA_MODELS[model]
 
+        self.client = None  # if self.client -> using ollama cloud
+        if model.endswith("cloud"):
+            self.client = Client(
+                host="https://ollama.com",
+                headers={"Authorization": "Bearer " + os.environ.get("OLLAMA_API_KEY")},
+            )
+
         self.model_id = self.model_config["id"]
-        self.client = httpx.Client(base_url=base_url, timeout=120.0)
 
     def complete(
         self,
@@ -85,20 +93,12 @@ def complete(
         temperature: float = 0.0,
     ) -> CompletionResult:
         try:
-            response = self.client.post(
-                "/api/generate",
-                json={
-                    "model": self.model_id,
-                    "prompt": prompt,
-                    "stream": False,
-                    "options": {
-                        "temperature": temperature,
-                        "num_predict": max_tokens,
-                    },
-                },
+            data = generate(
+                model=self.model_id,
+                prompt=prompt,
+                stream=False,
+                options={"temperature": temperature, "num_predict": max_tokens},
             )
-            response.raise_for_status()
-            data = response.json()
 
             return CompletionResult(
                 success=True,
@@ -112,13 +112,6 @@ def complete(
                 },
             )
 
-        except httpx.ConnectError:
-            return CompletionResult(
-                success=False,
-                content="",
-                model=self.model_id,
-                error=f"Cannot connect to Ollama at {self.base_url}. Is Ollama running?",
-            )
         except Exception as e:
             return CompletionResult(
                 success=False,
@@ -146,41 +139,39 @@ def complete_vision(
         try:
             image_b64 = base64.standard_b64encode(image_data).decode("utf-8")
 
-            response = self.client.post(
-                "/api/generate",
-                json={
-                    "model": self.model_id,
-                    "prompt": prompt,
-                    "images": [image_b64],
-                    "stream": False,
-                    "options": {
-                        "temperature": temperature,
-                        "num_predict": max_tokens,
-                    },
-                },
-            )
-            response.raise_for_status()
-            data = response.json()
+            if self.client:
+                data = self.client.generate(
+                    model=self.model,
+                    prompt=prompt,
+                    images=[image_b64],
+                    stream=False,
+                    options={"temperature": temperature, "num_predict": max_tokens},
+                )
+            else:
+                data = generate(
+                    model=self.model,
+                    prompt=prompt,
+                    images=[image_b64],
+                    stream=False,
+                    options={"temperature": temperature, "num_predict": max_tokens},
+                )
 
             return CompletionResult(
                 success=True,
                 content=data.get("response", ""),
-                input_tokens=data.get("prompt_eval_count", 0),
-                output_tokens=data.get("eval_count", 0),
+                input_tokens=0
+                if data.get("prompt_eval_count") is None
+                else data["prompt_eval_count"],
+                output_tokens=0
+                if data.get("eval_count") is None
+                else data["eval_count"],
                 model=self.model_id,
                 metadata={
                     "total_duration": data.get("total_duration"),
                     "vision": True,
                 },
             )
 
-        except httpx.ConnectError:
-            return CompletionResult(
-                success=False,
-                content="",
-                model=self.model_id,
-                error=f"Cannot connect to Ollama at {self.base_url}. Is Ollama running?",
-            )
         except Exception as e:
             return CompletionResult(
                 success=False,
@@ -210,9 +201,7 @@ def get_provider_name(cls) -> str:
     def list_local_models(self) -> list[str]:
         """List models available in the local Ollama installation."""
         try:
-            response = self.client.get("/api/tags")
-            response.raise_for_status()
-            data = response.json()
+            data = list_models()
             return [m["name"] for m in data.get("models", [])]
         except Exception:
             return []

src/harvestor/schemas/base.py

@@ -188,7 +188,7 @@ def to_summary(self) -> str:
         status = "SUCCESS" if self.success else "FAILED"
         strategy = self.final_strategy.value if self.final_strategy else "N/A"
 
-        return f"""
+        summary = f"""
             Harvest Result: {status}
             Document: {self.document_id} ({self.document_type})
             Strategy: {strategy}
@@ -199,6 +199,13 @@ def to_summary(self) -> str:
             Data: {self.data}
         """.strip()
 
+        if self.error:
+            summary += f"""
+            Error: {self.error}
+        """
+
+        return summary
+
 
 @dataclass
 class CostReport: