versionhq 1.1.11.5__tar.gz → 1.1.11.7__tar.gz

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/.github/workflows/run_tests.yml

@@ -8,13 +8,11 @@ permissions:
 env:
   LITELLM_API_KEY: ${{ secrets.LITELLM_API_KEY }}
   OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
-  OPENAI_API_BASE: ${{ secrets.OPENAI_API_BASE }}
   DEFAULT_REDIRECT_URL:  ${{ secrets.DEFAULT_REDIRECT_URL }}
   DEFAULT_USER_ID:  ${{ secrets.DEFAULT_USER_ID }}
   COMPOSIO_API_KEY: ${{ secrets.COMPOSIO_API_KEY }}
   DEFAULT_MODEL_NAME:  ${{ secrets.DEFAULT_MODEL_NAME }}
   GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
-  GEMINI_API_BASE: ${{ secrets.GEMINI_API_BASE }}
   ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
   ANTHROPIC_API_BASE: ${{ secrets.ANTHROPIC_API_BASE }}
   MEM0_API_KEY: ${{ secrets.MEM0_API_KEY }}

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: versionhq
-Version: 1.1.11.5
+Version: 1.1.11.7
 Summary: LLM orchestration frameworks for model-agnostic AI agents that handle complex outbound workflows
 Author-email: Kuriko Iwai <kuriko@versi0n.io>
 License: MIT License
@@ -319,8 +319,10 @@ src/
    pyenv install 3.12.8
    pyenv global 3.12.8  (optional: `pyenv global system` to get back to the system default ver.)
    uv python pin 3.12.8
+   echo 3.12.8 > .python-version
    ```
 
+
 3. Set up environment variables:
    Create a `.env` file in the project root and add the following:
    ```

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/README.md

@@ -243,8 +243,10 @@ src/
    pyenv install 3.12.8
    pyenv global 3.12.8  (optional: `pyenv global system` to get back to the system default ver.)
    uv python pin 3.12.8
+   echo 3.12.8 > .python-version
    ```
 
+
 3. Set up environment variables:
    Create a `.env` file in the project root and add the following:
    ```

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/pyproject.toml

@@ -15,7 +15,7 @@ exclude = ["test*", "__pycache__", "*.egg-info"]
 
 [project]
 name = "versionhq"
-version = "1.1.11.5"
+version = "1.1.11.7"
 authors = [{ name = "Kuriko Iwai", email = "kuriko@versi0n.io" }]
 description = "LLM orchestration frameworks for model-agnostic AI agents that handle complex outbound workflows"
 readme = "README.md"

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq/__init__.py

@@ -17,7 +17,7 @@ from versionhq.tool.model import Tool
 from versionhq.tool.composio_tool import ComposioHandler
 
 
-__version__ = "1.1.11.5"
+__version__ = "1.1.11.7"
 __all__ = [
     "Agent",
     "Customer",

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq/agent/model.py

@@ -99,7 +99,7 @@ class Agent(BaseModel):
     tools: Optional[List[InstanceOf[Tool | ToolSet] | Type[Tool] | Any]] = Field(default_factory=list)
 
     # knowledge
-    knowledge_sources: Optional[List[BaseKnowledgeSource]] = Field(default=None)
+    knowledge_sources: Optional[List[BaseKnowledgeSource | Any]] = Field(default=None)
     _knowledge: Optional[Knowledge] = PrivateAttr(default=None)
 
     # memory
@@ -344,14 +344,46 @@ class Agent(BaseModel):
 
     @model_validator(mode="after")
     def set_up_knowledge(self) -> Self:
-        if self.knowledge_sources:
-            collection_name = f"{self.role.replace(' ', '_')}"
+        from versionhq.knowledge.source import BaseKnowledgeSource, StringKnowledgeSource, TextFileKnowledgeSource, CSVKnowledgeSource, ExcelKnowledgeSource, JSONKnowledgeSource
+        from versionhq.knowledge.source_docling import DoclingSource
 
-            self._knowledge = Knowledge(
-                sources=self.knowledge_sources,
-                embedder_config=self.embedder_config,
-                collection_name=collection_name,
-            )
+        if self.knowledge_sources:
+            try:
+                collection_name = f"{self.role.replace(' ', '_')}"
+                knowledge_sources = []
+                docling_fp, txt_fp, json_fp, excel_fp, csv_fp, pdf_fp = [], [], [], [], [], []
+                str_cont = ""
+
+                for item in self.knowledge_sources:
+                    if isinstance(item, BaseKnowledgeSource):
+                        knowledge_sources.append(item)
+
+                    elif isinstance(item, str) and "http" in item:
+                        docling_fp.append(item)
+
+                    elif isinstance(item, str):
+                        match  os.path.splitext(item)[1]:
+                            case ".txt": txt_fp.append(item)
+                            case ".json": json_fp.append(item)
+                            case ".xls" | ".xlsx": excel_fp.append(item)
+                            case ".pdf": pdf_fp.append(item)
+                            case ".csv": csv_fp.append(item)
+                            case _: str_cont += str(item)
+
+                    else:
+                        str_cont += str(item)
+
+                if docling_fp: knowledge_sources.append(DoclingSource(file_paths=docling_fp))
+                if str_cont: knowledge_sources.append(StringKnowledgeSource(content=str_cont))
+                if txt_fp: knowledge_sources.append(TextFileKnowledgeSource(file_paths=txt_fp))
+                if csv_fp: knowledge_sources.append(CSVKnowledgeSource(file_path=csv_fp))
+                if excel_fp: knowledge_sources.append(ExcelKnowledgeSource(file_path=excel_fp))
+                if json_fp: knowledge_sources.append(JSONKnowledgeSource(file_paths=json_fp))
+
+                self._knowledge = Knowledge(sources=knowledge_sources, embedder_config=self.embedder_config, collection_name=collection_name)
+
+            except:
+                self._logger.log(level="warning", message="We cannot find the format for the source. Add BaseKnowledgeSource objects instead.", color="yellow")
 
         return self
 
@@ -414,7 +446,7 @@ class Agent(BaseModel):
             self._logger.log(level="info", message=f"Messages sent to the model: {messages}", color="blue")
 
             if tool_res_as_final:
-                func_llm = self.function_calling_llm if self.function_calling_llm and self.function_calling_llm._supports_function_calling() else LLM(model=DEFAULT_MODEL_NAME)
+                func_llm = self.function_calling_llm if self.function_calling_llm and self.function_calling_llm._supports_function_calling() else self.llm if self.llm and self.llm._supports_function_calling() else LLM(model=DEFAULT_MODEL_NAME)
                 raw_response = func_llm.call(messages=messages, tools=tools, tool_res_as_final=True)
                 task.tokens = func_llm._tokens
             else:
@@ -458,7 +490,7 @@ class Agent(BaseModel):
         from versionhq.knowledge._utils import extract_knowledge_context
 
         task: InstanceOf[Task] = task
-        tools: Optional[List[InstanceOf[Tool]| InstanceOf[ToolSet] | Type[Tool]]] = task_tools + self.tools if task.can_use_agent_tools else task_tools
+        tools: Optional[List[InstanceOf[Tool | ToolSet] | Type[Tool]]] = task_tools + self.tools if task.can_use_agent_tools else task_tools
 
         if self.max_rpm and self._rpm_controller:
             self._rpm_controller._reset_request_count()
@@ -474,7 +506,6 @@ class Agent(BaseModel):
                 if agent_knowledge_context:
                     task_prompt += agent_knowledge_context
 
-
         if self.use_memory == True:
             contextual_memory = ContextualMemory(
                 memory_config=self.memory_config, stm=self.short_term_memory, ltm=self.long_term_memory, um=self.user_memory

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq/knowledge/source_docling.py

@@ -3,7 +3,6 @@ from typing import Iterator, List, Optional
 from urllib.parse import urlparse
 
 try:
-    import docling
     from docling.datamodel.base_models import InputFormat
     from docling.document_converter import DocumentConverter
     from docling.exceptions import ConversionError
@@ -12,9 +11,8 @@ try:
     DOCLING_AVAILABLE = True
 except ImportError:
     import envoy
-    r = envoy.run("uv add docling --optional docling")
+    envoy.run("uv add docling --optional docling")
 
-    import docling
     from docling.datamodel.base_models import InputFormat
     from docling.document_converter import DocumentConverter
     from docling.exceptions import ConversionError

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq/knowledge/storage.py

@@ -73,11 +73,21 @@ class KnowledgeStorage(BaseKnowledgeStorage):
 
 
     def __init__(self, embedder_config: Optional[Dict[str, Any]] = None, collection_name: Optional[str] = None):
-        self.collection_name = collection_name if collection_name else "knowledge"
+        self.collection_name = self._validate_collection_name(collection_name) if collection_name else "knowledge"
         self.embedder_config = embedder_config
         self.initialize_knowledge_storage()
 
 
+    def _validate_collection_name(self, collection_name: str = None) -> str:
+        """
+        Return a valid collection name from the given collection name.
+        Expected collection name (1) contains 3-63 characters, (2) starts and ends with an alphanumeric character, (3) otherwise contains only alphanumeric characters, underscores or hyphens (-), (4) contains no two consecutive periods (..) and (5) is not a valid IPv4 address.
+        """
+        collection_name = collection_name if collection_name else self.collection_name
+        valid_collection_name = collection_name.replace(' ', "-").replace("(", "-").replace(")", "").replace("..", "")
+        return valid_collection_name
+
+
     def _create_default_embedding_function(self) -> Any:
         from chromadb.utils.embedding_functions.openai_embedding_function import OpenAIEmbeddingFunction
 
@@ -101,6 +111,7 @@ class KnowledgeStorage(BaseKnowledgeStorage):
         chroma_client = chromadb.PersistentClient(path=base_path, settings=Settings(allow_reset=True))
         self.app = chroma_client
         self._set_embedding_function(embedder_config=self.embedder_config)
+        self.collection_name = self.collection_name if self.collection_name else "knowledge"
 
         try:
             if self.app:

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq/llm/llm_vars.py

@@ -3,6 +3,20 @@ from typing import Type
 
 JSON_URL = "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
 
+PROVIDERS = [
+    "openai",
+    "gemini",
+    "sagemaker",
+
+    "anthropic",
+    "ollama",
+    "watson",
+    "bedrock",
+    "azure",
+    "cerebras",
+    "llama",
+]
+
 
 """
 List of models available on the framework.
@@ -16,7 +30,6 @@ litellm.pick_cheapest_chat_models_from_llm_provider(custom_llm_provider: str, n=
 
 MODELS = {
     "openai": [
-        # "gpt-3.5-turbo",
         "gpt-4",
         "gpt-4o",
         "gpt-4o-mini",
@@ -27,11 +40,7 @@ MODELS = {
         "gemini/gemini-1.5-flash",
         "gemini/gemini-1.5-pro",
         "gemini/gemini-2.0-flash-exp",
-        # "gemini/gemini-gemma-2-9b-it",
-        # "gemini/gemini-gemma-2-27b-it",
     ],
-    # "vetrex_ai": [
-    # ],
     "anthropic": [
         "claude-3-5-sonnet-20241022",
         "claude-3-5-sonnet-20240620",
@@ -39,10 +48,26 @@ MODELS = {
         "claude-3-opus-20240229",
         "claude-3-haiku-20240307",
     ],
-    # "ollama": [
-    #   "ollama/llama3.1",
-    #   "ollama/mixtral",
-    # ],
+    # "sagemaker": [
+    #     "sagemaker/huggingface-text2text-flan-t5-base",
+    #     "sagemaker/huggingface-llm-gemma-7b",
+    #     "sagemaker/jumpstart-dft-meta-textgeneration-llama-2-13b",
+    #     "sagemaker/jumpstart-dft-meta-textgeneration-llama-2-70b",
+    #     "sagemaker/jumpstart-dft-meta-textgeneration-llama-3-8b",
+    #     "sagemaker/jumpstart-dft-meta-textgeneration-llama-3-70b",
+    #     "sagemaker/huggingface-llm-mistral-7b"
+    # ], #https://docs.aws.amazon.com/sagemaker/latest/dg/jumpstart-foundation-models-latest.html
+    "ollama": [
+      "ollama/llama3.1",
+      "ollama/mixtral",
+      "ollama/mixtral-8x22B-Instruct-v0.1",
+
+    ],
+    "deepseek": [
+        "deepseek/deepseek-reasoner",
+
+    ],
+
     # "watson": [
     #     "watsonx/meta-llama/llama-3-1-70b-instruct",
     #     "watsonx/meta-llama/llama-3-1-8b-instruct",
@@ -53,44 +78,48 @@ MODELS = {
     #     "watsonx/mistral/mistral-large",
     #     "watsonx/ibm/granite-3-8b-instruct",
     # ],
-    # "bedrock": [
-    #     "bedrock/anthropic.claude-3-5-sonnet-20240620-v1:0",
-    #     "bedrock/anthropic.claude-3-sonnet-20240229-v1:0",
-    #     "bedrock/anthropic.claude-3-haiku-20240307-v1:0",
-    #     "bedrock/anthropic.claude-3-opus-20240229-v1:0",
-    #     "bedrock/anthropic.claude-v2:1",
-    #     "bedrock/anthropic.claude-v2",
-    #     "bedrock/anthropic.claude-instant-v1",
-    #     "bedrock/meta.llama3-1-405b-instruct-v1:0",
-    #     "bedrock/meta.llama3-1-70b-instruct-v1:0",
-    #     "bedrock/meta.llama3-1-8b-instruct-v1:0",
-    #     "bedrock/meta.llama3-70b-instruct-v1:0",
-    #     "bedrock/meta.llama3-8b-instruct-v1:0",
-    #     "bedrock/amazon.titan-text-lite-v1",
-    #     "bedrock/amazon.titan-text-express-v1",
-    #     "bedrock/cohere.command-text-v14",
-    #     "bedrock/ai21.j2-mid-v1",
-    #     "bedrock/ai21.j2-ultra-v1",
-    #     "bedrock/ai21.jamba-instruct-v1:0",
-    #     "bedrock/meta.llama2-13b-chat-v1",
-    #     "bedrock/meta.llama2-70b-chat-v1",
-    #     "bedrock/mistral.mistral-7b-instruct-v0:2",
-    #     "bedrock/mistral.mixtral-8x7b-instruct-v0:1",
-    # ],
+    "bedrock": [
+        "bedrock/anthropic.claude-3-5-sonnet-20240620-v1:0",
+        "bedrock/anthropic.claude-3-sonnet-20240229-v1:0",
+        "bedrock/anthropic.claude-3-haiku-20240307-v1:0",
+        "bedrock/anthropic.claude-3-opus-20240229-v1:0",
+        # "bedrock/anthropic.claude-v2:1",
+        "bedrock/anthropic.claude-v2",
+        "bedrock/anthropic.claude-instant-v1",
+        "bedrock/meta.llama3-1-405b-instruct-v1:0",
+        "bedrock/meta.llama3-1-70b-instruct-v1:0",
+        "bedrock/meta.llama3-1-8b-instruct-v1:0",
+        "bedrock/meta.llama3-70b-instruct-v1:0",
+        "bedrock/meta.llama3-8b-instruct-v1:0",
+        "bedrock/amazon.titan-text-lite-v1",
+        "bedrock/amazon.titan-text-express-v1",
+        "bedrock/cohere.command-text-v14",
+        "bedrock/ai21.j2-mid-v1",
+        "bedrock/ai21.j2-ultra-v1",
+        "bedrock/ai21.jamba-instruct-v1:0",
+        "bedrock/meta.llama2-13b-chat-v1",
+        "bedrock/meta.llama2-70b-chat-v1",
+        "bedrock/mistral.mistral-7b-instruct-v0:2",
+        "bedrock/mistral.mixtral-8x7b-instruct-v0:1",
+    ],
 }
 
 
-PROVIDERS = [
-    "openai",
-    "anthropic",
-    "gemini",
-    "ollama",
-    "watson",
-    "bedrock",
-    "azure",
-    "cerebras",
-    "llama",
-]
+
+KEYS = {
+    "openai": ["OPENAI_API_KEY"],
+    "gemini": ["GEMINI_API_KEY"],
+    "sagemaker": ["AWS_ACCESS_KEY_ID", "ADW_SECURET_ACCESS_KEY", "AWS_REGION_NAME"],
+    "anthropic": ["ANTHROPIC_API_KEY"],
+}
+
+
+"""
+Use base_url to specify
+"""
+BASE_URLS = {
+    "deepseek": "https://api.deepseek.com"
+}
 
 
 """
@@ -118,6 +147,8 @@ LLM_CONTEXT_WINDOW_SIZES = {
     "claude-3-haiku-20240307": 200000,
 
     "deepseek-chat": 128000,
+    "deepseek/deepseek-reasoner": 8192,
+
     "gemma2-9b-it": 8192,
     "gemma-7b-it": 8192,
     "llama3-groq-70b-8192-tool-use-preview": 8192,
@@ -135,11 +166,7 @@ LLM_CONTEXT_WINDOW_SIZES = {
 }
 
 
-LLM_API_KEY_NAMES = {
-    "openai":  "OPENAI_API_KEY",
-    "anthropic": "ANTHROPIC_API_KEY",
-    "gemini": "GEMINI_API_KEY",
-}
+
 
 LLM_BASE_URL_KEY_NAMES = {
     "openai":  "OPENAI_API_BASE",
@@ -262,14 +289,8 @@ PARAMS = {
     ],
     "openai": [
         "timeout",
-        # "temperature",
-        # "top_p",
-        # "n",
-        # "stream",
         "stream_options",
-        # "stop",
         "max_compl,etion_tokens",
-        # "max_tokens",
         "modalities",
         "prediction",
         "audio",
@@ -277,10 +298,7 @@ PARAMS = {
         "frequency_penalty",
         "logit_bias",
         "user",
-        # "response_format",
         "seed",
-        # "tools",
-        # "tool_choice",
         "logprobs",
         "top_logprobs",
         "parallel_tool_calls",

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq/llm/model.py

@@ -4,23 +4,16 @@ import os
 import sys
 import threading
 import warnings
-import litellm
-from litellm import JSONSchemaValidationError
-from abc import ABC
 from dotenv import load_dotenv
-from litellm import get_supported_openai_params
+import litellm
+from litellm import get_supported_openai_params, JSONSchemaValidationError
 from contextlib import contextmanager
-from typing import Any, Dict, List, Optional, Type
+from typing import Any, Dict, List, Optional
 from typing_extensions import Self
-
 from pydantic import UUID4, BaseModel, Field, PrivateAttr, field_validator, model_validator, create_model, InstanceOf, ConfigDict
 from pydantic_core import PydanticCustomError
 
-from openai import OpenAI
-
-from versionhq.llm.llm_vars import LLM_CONTEXT_WINDOW_SIZES, LLM_API_KEY_NAMES, LLM_BASE_URL_KEY_NAMES, MODELS, PARAMS, SchemaType
-from versionhq.task import TaskOutputFormat
-from versionhq.task.model import ResponseField, Task
+from versionhq.llm.llm_vars import LLM_CONTEXT_WINDOW_SIZES, MODELS, PARAMS
 from versionhq.tool.model import Tool, ToolSet
 from versionhq._utils.logger import Logger
 
@@ -31,8 +24,8 @@ LITELLM_API_BASE = os.environ.get("LITELLM_API_BASE")
 DEFAULT_CONTEXT_WINDOW_SIZE = int(8192 * 0.75)
 DEFAULT_MODEL_NAME = os.environ.get("DEFAULT_MODEL_NAME")
 
-proxy_openai_client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"), organization="versionhq", base_url=LITELLM_API_BASE)
-openai_client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))
+# proxy_openai_client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"), organization="versionhq", base_url=LITELLM_API_BASE)
+# openai_client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))
 
 
 class FilteredStream:
@@ -179,6 +172,7 @@ class LLM(BaseModel):
         if api_key_name:
             self.api_key = os.environ.get(api_key_name, None)
 
+
         base_url_key_name = self.provider.upper() + "_API_BASE" if self.provider else None
         if base_url_key_name:
             self.base_url = os.environ.get(base_url_key_name)
@@ -236,51 +230,51 @@ class LLM(BaseModel):
                 else:
                     self.tools = [item.tool.properties if isinstance(item, ToolSet) else item.properties for item in tools]
 
-                    if provider == "openai":
-                        params = self._create_valid_params(config=config, provider=provider)
-                        res = openai_client.chat.completions.create(messages=messages, model=self.model, tools=self.tools)
-                        tool_calls = res.choices[0].message.tool_calls
-                        tool_res = ""
-
-                        for item in tool_calls:
-                            func_name = item.function.name
-                            func_args = item.function.arguments
+                    # if provider == "openai":
+                    params = self._create_valid_params(config=config, provider=provider)
+                    res = litellm.completion(messages=messages, model=self.model, tools=self.tools)
+                    tool_calls = res.choices[0].message.tool_calls
+                    tool_res = ""
+
+                    for item in tool_calls:
+                        func_name = item.function.name
+                        func_args = item.function.arguments
+
+                        if not isinstance(func_args, dict):
+                            try:
+                                func_args = json.loads(json.dumps(eval(str(func_args))))
+                            except:
+                                pass
+
+                        for tool in tools:
+                            if isinstance(tool, ToolSet) and (tool.tool.name == func_name or tool.tool.func.__name__ == func_name or func_name == "random_func"):
+                                tool_instance = tool.tool
+                                args = tool.kwargs
+                                tool_res_to_add = tool_instance.run(params=args)
+
+                                if tool_res_as_final:
+                                    tool_res += str(tool_res_to_add)
+                                else:
+                                    messages.append(res.choices[0].message)
+                                    messages.append({ "role": "tool", "tool_call_id": item.id, "content": str(tool_res_to_add) })
 
-                            if not isinstance(func_args, dict):
+                            else:
                                 try:
-                                    func_args = json.loads(json.dumps(eval(str(func_args))))
-                                except:
-                                    pass
-
-                            for tool in tools:
-                                if isinstance(tool, ToolSet) and (tool.tool.name == func_name or tool.tool.func.__name__ == func_name or func_name == "random_func"):
-                                    tool_instance = tool.tool
-                                    args = tool.kwargs
-                                    tool_res_to_add = tool_instance.run(params=args)
-
+                                    tool_res_to_add = tool.run(params=func_args)
                                     if tool_res_as_final:
                                         tool_res += str(tool_res_to_add)
                                     else:
                                         messages.append(res.choices[0].message)
                                         messages.append({ "role": "tool", "tool_call_id": item.id, "content": str(tool_res_to_add) })
+                                except:
+                                    pass
 
-                                else:
-                                    try:
-                                        tool_res_to_add = tool.run(params=func_args)
-                                        if tool_res_as_final:
-                                            tool_res += str(tool_res_to_add)
-                                        else:
-                                            messages.append(res.choices[0].message)
-                                            messages.append({ "role": "tool", "tool_call_id": item.id, "content": str(tool_res_to_add) })
-                                    except:
-                                        pass
-
-                        if tool_res_as_final:
-                            return tool_res
-                        else:
-                            res = openai_client.chat.completions.create(messages=messages, model=self.model, tools=self.tools)
-                            self._tokens += int(res["usage"]["total_tokens"])
-                            return res.choices[0].message.content
+                    if tool_res_as_final:
+                        return tool_res
+                    else:
+                        res = litellm.completione(messages=messages, model=self.model, tools=self.tools)
+                        self._tokens += int(res["usage"]["total_tokens"])
+                        return res.choices[0].message.content
 
             except JSONSchemaValidationError as e:
                 self._logger.log(level="error", message="Raw Response: {}".format(e.raw_response), color="red")

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq/task/model.py

@@ -412,6 +412,8 @@ Ref. Output image: {output_formats_to_follow}
 
         response_format: Dict[str, Any] = None
 
+        # match model_provider:
+        #     case "openai":
         if self.response_fields:
             properties, required_fields = {}, []
             for i, item in enumerate(self.response_fields):
@@ -439,6 +441,7 @@ Ref. Output image: {output_formats_to_follow}
         elif self.pydantic_output:
             response_format = StructuredOutput(response_format=self.pydantic_output)._format()
 
+            # case "gemini":
         return response_format
 
 
@@ -636,7 +639,7 @@ Ref. Output image: {output_formats_to_follow}
 
         if self.tool_res_as_final == True:
             tool_output = agent.execute_task(task=self, context=context, task_tools=task_tools)
-            task_output = TaskOutput(task_id=self.id, tool_output=tool_output)
+            task_output = TaskOutput(task_id=self.id, tool_output=tool_output, raw=tool_output)
 
         else:
             raw_output = agent.execute_task(task=self, context=context, task_tools=task_tools)

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq/task/structured_response.py

@@ -82,12 +82,13 @@ class StructuredList:
 
                 if nested_object_type == dict:
                     props.update({
-                        "nest":  {
+                        # "nest":  {
                             "type": "object",
                             "properties": { "item": { "type": "string"} }, #! REFINEME - field title <>`item`
                             "required": ["item",],
                             "additionalProperties": False
-                        }})
+                        # }
+                        })
 
                 elif nested_object_type == list:
                     props.update({
@@ -110,7 +111,7 @@ class StructuredList:
 
 
 class StructuredOutput(BaseModel):
-    response_format: Any = None
+    response_format: Any = None # pydantic base model
     provider: str = "openai"
     applicable_models: List[InstanceOf[LLM] | str] = list()
     name: str = ""

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/src/versionhq.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.2
 Name: versionhq
-Version: 1.1.11.5
+Version: 1.1.11.7
 Summary: LLM orchestration frameworks for model-agnostic AI agents that handle complex outbound workflows
 Author-email: Kuriko Iwai <kuriko@versi0n.io>
 License: MIT License
@@ -319,8 +319,10 @@ src/
    pyenv install 3.12.8
    pyenv global 3.12.8  (optional: `pyenv global system` to get back to the system default ver.)
    uv python pin 3.12.8
+   echo 3.12.8 > .python-version
    ```
 
+
 3. Set up environment variables:
    Create a `.env` file in the project root and add the following:
    ```

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/tests/agent/agent_test.py

@@ -231,22 +231,26 @@ def test_agent_custom_max_iterations():
 
 def test_agent_with_knowledge_sources():
     from versionhq.knowledge.source import StringKnowledgeSource
+    from versionhq.knowledge.source_docling import DoclingSource
     from versionhq.task.model import Task
 
     content = "Kuriko's favorite color is gold, and she enjoy Japanese food."
     string_source = StringKnowledgeSource(content=content)
+    html = "https://github.blog/security/vulnerability-research/cybersecurity-researchers-digital-detectives-in-a-connected-world/"
+    knowledge_sources = [content, string_source, html,]
 
-    agent = Agent(role="Information Agent", goal="Provide information based on knowledge sources", knowledge_sources=[string_source])
+    agent = Agent(role="Information Agent", goal="Provide information based on knowledge sources", knowledge_sources=knowledge_sources)
 
     assert agent._knowledge.collection_name == f"{agent.role.replace(' ', '_')}"
-    assert agent._knowledge.sources == [string_source] and agent._knowledge.embedder_config == agent.embedder_config
+    assert [isinstance(item, StringKnowledgeSource | DoclingSource) for item in agent.knowledge_sources]
+    assert agent._knowledge.embedder_config == agent.embedder_config
     assert agent._knowledge.storage and agent._knowledge.storage.embedding_function and  agent._knowledge.storage.app is not None and agent._knowledge.storage.collection_name is not None
 
     task = Task(description="Answer the following question: What is Kuriko's favorite color?")
 
     with patch("versionhq.knowledge.storage.KnowledgeStorage") as MockKnowledge:
         mock_knowledge_instance = MockKnowledge.return_value
-        mock_knowledge_instance.sources = [string_source, ]
+        isinstance(mock_knowledge_instance.sources[0], StringKnowledgeSource)
         mock_knowledge_instance.query.return_value = [{ "content": content }]
 
         res = task.execute_sync(agent=agent)
@@ -308,3 +312,8 @@ def test_agent_with_memory_config():
     assert agent_2.short_term_memory.memory_provider == "mem0" and agent_2.short_term_memory.storage.memory_type == "stm"
     assert agent_2.long_term_memory and isinstance(agent_2.long_term_memory.storage, LTMSQLiteStorage)
     assert agent_2.user_memory and agent_2.user_memory.storage and agent_2.user_memory.storage.memory_type == "user"
+
+
+
+if __name__ == "__main__":
+    test_agent_with_knowledge_sources()

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/tests/llm/llm_test.py

@@ -7,7 +7,7 @@ from versionhq.llm.llm_vars import MODELS, LLM_CONTEXT_WINDOW_SIZES
 from versionhq.llm.model import LLM, DEFAULT_CONTEXT_WINDOW_SIZE, DEFAULT_MODEL_NAME
 
 
-def _dummy_func() -> str:
+def dummy_func() -> str:
     return "dummy"
 
 
@@ -18,49 +18,39 @@ def test_create_llm_from_valid_name():
 
     for k, v in MODELS.items():
         for model_name in v:
-            llm = LLM(model=model_name, callbacks=[_dummy_func,])
+            llm = LLM(model=model_name, callbacks=[dummy_func,])
 
             assert llm._init_model_name == model_name
             assert llm.model == model_name
             assert llm.provider == k
-            assert llm.api_key is not None
-            assert llm.base_url is not None
             assert llm.context_window_size == int(LLM_CONTEXT_WINDOW_SIZES.get(model_name) *0.75) if LLM_CONTEXT_WINDOW_SIZES.get(model_name) is not None else DEFAULT_CONTEXT_WINDOW_SIZE
             assert llm._supports_function_calling() is not None
             assert llm._supports_stop_words() is not None
-            assert litellm.callbacks == [_dummy_func,]
+            assert litellm.callbacks == [dummy_func,]
 
 
 def test_create_llm_from_invalid_name():
     """
     Test if all the params will be set properly with a givne invalid model name.
     """
-    llm = LLM(model="4o", callbacks=[_dummy_func,])
+    llm = LLM(model="4o", callbacks=[dummy_func,])
 
     assert llm._init_model_name == "4o"
     assert llm.model == "gpt-4o"
     assert llm.provider == "openai"
-    assert llm.api_key is not None
-    assert llm.base_url is not None
     assert llm.context_window_size == int(128000 * 0.75)
     assert llm._supports_function_calling() == True
     assert llm._supports_stop_words() == True
-    assert litellm.callbacks == [_dummy_func,]
+    assert litellm.callbacks == [dummy_func,]
 
 
 def test_create_llm_from_provider():
-    llm = LLM(provider="gemini", callbacks=[_dummy_func,])
+    llm = LLM(provider="gemini", callbacks=[dummy_func,])
 
     assert llm._init_model_name == DEFAULT_MODEL_NAME
     assert llm.model == "gemini/gemini-1.5-flash"
     assert llm.provider == "gemini"
-    assert llm.api_key is not None
-    assert llm.base_url is not None
     assert llm.context_window_size == int(LLM_CONTEXT_WINDOW_SIZES.get(llm.model) *0.75) if LLM_CONTEXT_WINDOW_SIZES.get(llm.model) is not None else DEFAULT_CONTEXT_WINDOW_SIZE
     assert llm._supports_function_calling() == True
     assert llm._supports_stop_words() == True
-    assert litellm.callbacks == [_dummy_func,]
-
-
-if __name__ == "__main__":
-    test_create_llm_from_valid_name()
+    assert litellm.callbacks == [dummy_func,]

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/tests/task/task_test.py

@@ -350,7 +350,6 @@ def test_evaluation():
     """
     See if the output will be evaluated accurately - when the task was given eval criteria
     """
-    from versionhq.task.model import Task
     from versionhq.task.evaluate import Evaluation, EvaluationItem
     from versionhq.agent.default_agents import task_evaluator
 
@@ -366,3 +365,54 @@ def test_evaluation():
     assert [isinstance(item, EvaluationItem) and item.criteria in task.eval_criteria for item in res.evaluation.items]
     assert res.evaluation.latency and res.evaluation.tokens and res.evaluation.responsible_agent == task_evaluator
     assert res.evaluation.aggregate_score is not None and res.evaluation.suggestion_summary
+
+
+
+def test_gemini_schema():
+    """
+    See if response schema and tools (func_calling) works.
+    """
+    from tests.task import DemoOutcome
+    agent = Agent(role="demo", goal="demo", llm="gemini/gemini-1.5-pro")
+    task = Task(
+        description="return random values strictly following the given response format.",
+        pydantic_output=DemoOutcome
+    )
+    res = task.execute_sync(agent=agent, context="We are running a test.")
+    assert [
+        getattr(res.pydantic, k) and type(getattr(res.pydantic, k)) == v for k, v in DemoOutcome.__annotations__.items()
+    ]
+
+
+def test_gemini_res_fields():
+    from tests.task import demo_response_fields
+    agent = Agent(role="demo", goal="demo", llm="gemini/gemini-1.5-pro")
+    task = Task(
+        description="return random values strictly following the given response format.",
+        response_fields=demo_response_fields
+    )
+    res = task.execute_sync(agent=agent, context="We are running a test.")
+    assert [k in item.title for item in demo_response_fields for k, v in res.json_dict.items()]
+
+
+def test_gemini_func():
+    from tests.task import demo_response_fields
+    from versionhq.tool.model import Tool
+
+    class DemoTool(Tool):
+        func: Callable[..., Any] = lambda x: "Gemini"
+
+    agent = Agent(role="demo", goal="demo", llm="gemini/gemini-1.5-pro")
+    task = Task(
+        description="Simply execute the given tools.",
+        tools=[DemoTool,],
+        tool_res_as_final=True
+    )
+    res = task.execute_sync(agent=agent, context="We are running a test.")
+    assert res.tool_output and res.raw
+
+
+if __name__ == "__main__":
+    from dotenv import load_dotenv
+    load_dotenv(override=True)
+    test_gemini_func()

{versionhq-1.1.11.5 → versionhq-1.1.11.7}/uv.lock

@@ -1096,6 +1096,7 @@ dependencies = [
     { name = "hpack" },
     { name = "hyperframe" },
 ]
+sdist = { url = "https://files.pythonhosted.org/packages/1b/38/d7f80fd13e6582fb8e0df8c9a653dcc02b03ca34f4d72f34869298c5baf8/h2-4.2.0.tar.gz", hash = "sha256:c8a52129695e88b1a0578d8d2cc6842bbd79128ac685463b887ee278126ad01f", size = 2150682 }
 wheels = [
     { url = "https://files.pythonhosted.org/packages/d0/9e/984486f2d0a0bd2b024bf4bc1c62688fcafa9e61991f041fb0e2def4a982/h2-4.2.0-py3-none-any.whl", hash = "sha256:479a53ad425bb29af087f3458a61d30780bc818e4ebcf01f0b536ba916462ed0", size = 60957 },
 ]
@@ -4445,7 +4446,7 @@ wheels = [
 
 [[package]]
 name = "versionhq"
-version = "1.1.11.5"
+version = "1.1.11.7"
 source = { editable = "." }
 dependencies = [
     { name = "appdirs" },