ollama-haystack 2.4.2__tar.gz → 3.0.0__tar.gz

{ollama_haystack-2.4.2 → ollama_haystack-3.0.0}/CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## [integrations/ollama-v2.4.2] - 2025-05-07
+
+### 🐛 Bug Fixes
+
+- Ollama streaming metadata 1686 (#1698)
+
+
 ## [integrations/ollama-v2.4.1] - 2025-04-10
 
 ### 🚀 Features

{ollama_haystack-2.4.2 → ollama_haystack-3.0.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: ollama-haystack
-Version: 2.4.2
+Version: 3.0.0
 Summary: An integration between the Ollama LLM framework and Haystack
 Project-URL: Documentation, https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/ollama#readme
 Project-URL: Issues, https://github.com/deepset-ai/haystack-core-integrations/issues
@@ -19,7 +19,7 @@ Classifier: Programming Language :: Python :: 3.13
 Classifier: Programming Language :: Python :: Implementation :: CPython
 Classifier: Programming Language :: Python :: Implementation :: PyPy
 Requires-Python: >=3.9
-Requires-Dist: haystack-ai
+Requires-Dist: haystack-ai>=2.13.1
 Requires-Dist: ollama>=0.4.0
 Requires-Dist: pydantic
 Description-Content-Type: text/markdown

{ollama_haystack-2.4.2 → ollama_haystack-3.0.0}/pyproject.toml

@@ -27,7 +27,7 @@ classifiers = [
     "Programming Language :: Python :: Implementation :: CPython",
     "Programming Language :: Python :: Implementation :: PyPy",
 ]
-dependencies = ["haystack-ai", "ollama>=0.4.0", "pydantic"]
+dependencies = ["haystack-ai>=2.13.1", "ollama>=0.4.0", "pydantic"]
 
 [project.urls]
 Documentation = "https://github.com/deepset-ai/haystack-core-integrations/tree/main/integrations/ollama#readme"

{ollama_haystack-2.4.2 → ollama_haystack-3.0.0}/src/haystack_integrations/components/generators/ollama/chat/chat_generator.py

@@ -2,15 +2,8 @@ from typing import Any, Callable, Dict, List, Literal, Optional, Union
 
 from haystack import component, default_from_dict, default_to_dict
 from haystack.dataclasses import ChatMessage, StreamingChunk, ToolCall
-from haystack.tools import Tool, _check_duplicate_tool_names
+from haystack.tools import Tool, _check_duplicate_tool_names, deserialize_tools_or_toolset_inplace
 from haystack.utils.callable_serialization import deserialize_callable, serialize_callable
-
-# Compatibility with Haystack 2.12.0 and 2.13.0 - remove after 2.13.0 is released
-try:
-    from haystack.tools import deserialize_tools_or_toolset_inplace
-except ImportError:
-    from haystack.tools import deserialize_tools_inplace as deserialize_tools_or_toolset_inplace
-
 from pydantic.json_schema import JsonSchemaValue
 
 from ollama import ChatResponse, Client

{ollama_haystack-2.4.2 → ollama_haystack-3.0.0}/src/haystack_integrations/components/generators/ollama/generator.py

@@ -7,6 +7,73 @@ from haystack.utils.callable_serialization import deserialize_callable, serializ
 from ollama import Client, GenerateResponse
 
 
+def _convert_ollama_meta_to_openai_format(intput_response_dict: Dict) -> Dict:
+    """
+    Converts Ollama metadata format to OpenAI-metadata format for standardization.
+
+    The conversion follows these mapping rules:
+    - 'done_reason' -> 'finish_reason'
+    - 'created_at' -> 'completion_start_time'
+    - 'eval_count' and 'prompt_eval_count' -> 'usage' dict with token counts
+    - All other fields are preserved as-is except 'response'.
+
+    Input Ollama-metadata format example:
+    {
+        'model': 'llama3.2:3b',
+        'created_at': '2025-05-15T14:11:57.8920338Z',
+        'done': True,
+        'done_reason': 'stop',
+        'total_duration': 871519900,
+        'load_duration': 44376100,
+        'prompt_eval_count': 29,
+        'prompt_eval_duration': 380700800,
+        'eval_count': 8,
+        'eval_duration': 445131300,
+        'context': [28006,
+         9125,
+         ...
+         220,
+         18]
+    }
+
+    Returned OpenAI-metadata format example:
+    {
+        'model': 'llama3.2:3b',
+        'finish_reason': 'stop',
+        'usage': {
+            'completion_tokens': 8,
+            'prompt_tokens': 29,
+            'total_tokens': 37,
+        }
+        'created_at': '2025-05-15T14:11:57.8920338Z',
+        'done': True,
+        'total_duration': 871519900,
+        'load_duration': 44376100,
+        'prompt_eval_duration': 380700800,
+        'eval_duration': 445131300,
+        'context': [28006,
+         9125,
+         ...
+         220,
+         18]
+    }
+    """
+
+    meta = {key: value for key, value in intput_response_dict.items() if key != "response"}
+
+    if "done_reason" in meta:
+        meta["finish_reason"] = meta.pop("done_reason")
+    if "eval_count" in meta and "prompt_eval_count" in meta:
+        eval_count = meta.pop("eval_count")
+        prompt_eval_count = meta.pop("prompt_eval_count")
+        meta["usage"] = {
+            "completion_tokens": eval_count,
+            "prompt_tokens": prompt_eval_count,
+            "total_tokens": eval_count + prompt_eval_count,
+        }
+    return meta
+
+
 @component
 class OllamaGenerator:
     """
@@ -122,8 +189,9 @@ class OllamaGenerator:
         """
         Converts a response from the Ollama API to the required Haystack format.
         """
-        reply = ollama_response.response
-        meta = {key: value for key, value in ollama_response.model_dump().items() if key != "response"}
+        response_dict = ollama_response.model_dump()
+        reply = response_dict["response"]
+        meta = _convert_ollama_meta_to_openai_format(response_dict)
 
         return {"replies": [reply], "meta": [meta]}
 
@@ -133,7 +201,9 @@ class OllamaGenerator:
         """
 
         replies = ["".join([c.content for c in chunks])]
-        meta = {key: value for key, value in chunks[0].meta.items() if key != "response"}
+
+        # Convert the metadata from the last chunk
+        meta = _convert_ollama_meta_to_openai_format(chunks[-1].meta)
 
         return {"replies": replies, "meta": [meta]}
 

{ollama_haystack-2.4.2 → ollama_haystack-3.0.0}/tests/test_chat_generator.py

@@ -574,10 +574,11 @@ class TestOllamaChatGenerator:
         response_format = {
             "type": "object",
             "properties": {"capital": {"type": "string"}, "population": {"type": "number"}},
+            "required": ["capital", "population"],
         }
         chat_generator = OllamaChatGenerator(model="llama3.2:3b", response_format=response_format)
 
-        message = ChatMessage.from_user("What's the capital of France and its population?")
+        message = ChatMessage.from_user("What's the capital of France and its population? Respond in JSON format.")
         response = chat_generator.run([message])
 
         assert isinstance(response, dict)