--- title: "Python LLM Adapters" description: "Python language model adapters power the memory extraction, context evaluation, and tool execution pipelines across the memory engine. By providing a uniform interface through Memory Lifecycle inte..." last_updated: "2026-10-11T17:37:10.25946+00:00" canonical_url: "https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/llm-providers/python-llm-adapters" ---
Relevant source files The following files were used as context for generating this wiki page: - [mem0-ts/src/oss/src/utils/factory.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/utils/factory.ts) - [mem0/llms/aws_bedrock.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/aws_bedrock.py) - [integrations/hermes-plugin-mem0/_backend.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/integrations/hermes-plugin-mem0/_backend.py) - [mem0/utils/factory.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/utils/factory.py) - [mem0/llms/deepseek.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/deepseek.py) - [mem0/memory/main.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/memory/main.py) - [mem0/llms/base.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/base.py) - [mem0/llms/__init__.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/__init__.py) - [mem0-ts/src/oss/src/llms/aws_bedrock.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/llms/aws_bedrock.ts) - [integrations/hermes-plugin-mem0/__init__.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/integrations/hermes-plugin-mem0/__init__.py) - [integrations/openclaw/cli/oss-wizard.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/integrations/openclaw/cli/oss-wizard.ts) - [mem0/llms/groq.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/groq.py) - [mem0-ts/src/oss/src/llms/langchain.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/llms/langchain.ts) - [integrations/kimi-plugin/hooks/adapter.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/integrations/kimi-plugin/hooks/adapter.py) - [mem0/llms/gemini.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/gemini.py) - [mem0/llms/lmstudio.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/lmstudio.py) - [integrations/hermes-plugin-mem0/_openai_llm.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/integrations/hermes-plugin-mem0/_openai_llm.py) - [mem0/configs/llms/__init__.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/configs/llms/__init__.py) - [mem0/llms/azure_openai_structured.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/azure_openai_structured.py) - [mem0/llms/ollama.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/ollama.py) - [mem0-ts/src/oss/src/llms/deepseek.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/llms/deepseek.ts) - [mem0/llms/xai.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/xai.py) - [mem0/llms/openai_structured.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/openai_structured.py) - [mem0-ts/src/oss/src/llms/ollama.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/llms/ollama.ts) - [mem0/llms/openai.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/openai.py) - [mem0/llms/langchain.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/langchain.py) - [mem0/llms/litellm.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/litellm.py) - [mem0/llms/vllm.py](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/vllm.py) - [mem0-ts/src/oss/src/llms/openai_structured.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/llms/openai_structured.ts) - [mem0-ts/src/oss/src/llms/groq.ts](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0-ts/src/oss/src/llms/groq.ts)
## Overview Python language model adapters power the memory extraction, context evaluation, and tool execution pipelines across the memory engine. By providing a uniform interface through [Memory Lifecycle](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/python-core/memory-lifecycle) integrations, these adapters handle parameter filtering, reasoning model detection, and provider-specific payload formatting for diverse backend providers. Sources: [mem0/llms/base.py:7-11](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/base.py#L7-L11), [mem0/memory/main.py:958-964](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/memory/main.py#L958-L964) ## LLMBase abstraction and lifecycle ### Parameter Filtering and Reasoning Model Detection The `LLMBase` abstract base class provides the foundational interface and configuration lifecycle for all language model providers in the engine. When an adapter is instantiated, the `__init__` method accepts either a `BaseLlmConfig` instance or a raw configuration dictionary, falling back to a default `BaseLlmConfig` if no argument is provided, and subsequently executes `_validate_config` to guarantee that the required `model` attribute is present on the configuration object. Sources: [mem0/llms/base.py:7-37](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/base.py#L7-L37) The base class implements name-based heuristics and explicit configuration overrides to classify reasoning models and adjust parameter payloads accordingly via `_is_reasoning_model` and `_get_supported_params`. An explicit `is_reasoning_model` boolean on the configuration takes precedence over string matching, allowing custom or versioned model deployments to opt in or out directly. When this configuration attribute is `None`, the model name is evaluated against a predefined set of base identifiers and prefix families. > [!NOTE] > The reasoning classification strips provider prefixes (such as `openai/o3-mini`) using `rsplit("/", 1)[-1]` before matching against the base set. The `o1` and `o3` family prefixes match variants such as `o1-` and `o3-`, whereas `gpt-5.x` variants like `gpt-5.4-mini` are explicitly excluded from the reasoning prefix match so that they retain standard parameter support. Sources: [mem0/llms/base.py:38-75](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/base.py#L38-L75) ```python reasoning_models = { "o1", "o1-preview", "o3-mini", "o3", "gpt-5", "gpt-5o", "gpt-5o-mini", "gpt-5o-micro", } model_lower = model.lower() # Strip provider prefixes (e.g. "openai/o3-mini" -> "o3-mini") base_model = model_lower.rsplit("/", 1)[-1] if base_model in reasoning_models: return True # Match o1/o3 family with prefixes (o1-2024-12-17, o3-2025-04-16) # but NOT gpt-5.x variants (gpt-5.4-mini supports temperature) if any(base_model.startswith(prefix) for prefix in ["o1-", "o1.", "o3-", "o3."]): return True return False ``` Sources: [mem0/llms/base.py:58-75](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/base.py#L58-L75) When parameter filtering executes in `_get_supported_params`, reasoning models and the `gpt-5` family restrict outbound parameters to a strict subset of keys, while regular models incorporate common parameters such as `temperature`, `top_p`, and token limits through `_get_common_params`. Furthermore, `_uses_max_completion_tokens` checks whether the model starts with `gpt-5` to substitute `max_tokens` with `max_completion_tokens` on the chat completions payload. | Model Category | Included Parameters / Keys | Token Limit Parameter | | :--- | :--- | :--- | | **Reasoning Models** (`o1`, `o3`, `gpt-5` base) | `messages`, `response_format`, `tools`, `tool_choice`, `reasoning_effort` | *None* | | **Regular Models** (`gpt-4.x`, `gpt-3.5`, etc.) | `temperature`, `top_p`, plus provider `kwargs` | `max_tokens` or `max_completion_tokens` | Sources: [mem0/llms/base.py:77-170](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/base.py#L77-L170) ## Structured output adapters ### Native OpenAI and Azure OpenAI Structured Classes The structured output adapters implement strict schema parsing and specialized API interactions for OpenAI and Azure OpenAI deployments. `OpenAIStructuredLLM` inherits from `LLMBase` and initializes an OpenAI client using `api_key` and `base_url` resolution, falling back to `"gpt-5-mini"` if the configuration model is empty. Its `generate_response` method delegates payload assembly to `_get_supported_params`, injects `response_format` and tool parameters when present, and executes parsed extractions using `client.beta.chat.completions.parse`. Sources: [mem0/llms/openai_structured.py:10-49](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/openai_structured.py#L10-L49) `AzureOpenAIStructuredLLM` provides an equivalent interface tailored for Azure deployments, resolving credentials through `DefaultAzureCredential` and bearer token providers when the `api_key` is absent, empty, or set to the `"your-api-key"` placeholder. It manages Azure endpoints, deployments, and API versions through `self.config.azure_kwargs` or matching environment variables (`LLM_AZURE_OPENAI_API_KEY`, `LLM_AZURE_DEPLOYMENT`, `LLM_AZURE_ENDPOINT`, `LLM_AZURE_API_VERSION`). > [!WARNING] > Azure's indirect attack content filters can flag the literal word `assistant` in prompt message histories. `AzureOpenAIStructuredLLM` automatically runs `_rewrite_assistant_keyword` on deep-copied messages before sending requests, replacing `"assistant"` with `"ai"` exclusively in string-based content properties while leaving multimodal list structures untouched. Sources: [mem0/llms/azure_openai_structured.py:16-124](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/azure_openai_structured.py#L16-L124) ```python @staticmethod def _rewrite_assistant_keyword(messages): """ Return a copy of ``messages`` with the word "assistant" replaced by "ai" in the last message's textual content. Azure's content management policy can flag the literal word "assistant", which makes ``add`` fail (see issue #2636). The rewrite targets that trigger without mutating the caller's messages and without assuming the content is a string, so multimodal (list) content passes through untouched. """ if not messages: return messages messages = copy.deepcopy(messages) last_content = messages[-1].get("content") if isinstance(last_content, str): messages[-1]["content"] = last_content.replace("assistant", "ai") return messages ``` Sources: [mem0/llms/azure_openai_structured.py:105-124](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/azure_openai_structured.py#L105-L124) ### Parameter Formatting and Response Parsing When generating responses under `AzureOpenAIStructuredLLM`, reasoning models bypass standard sampling parameters like `temperature` and `top_p`, while `max_tokens` is dynamically mapped to `max_completion_tokens` for reasoning models and `gpt-5` series variants. Optional `reasoning_effort` configurations are appended when specified on the configuration object. Sources: [mem0/llms/azure_openai_structured.py:76-101](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/azure_openai_structured.py#L76-L101) | Parameter Key | Reasoning / GPT-5 Models | Standard Models | | :--- | :--- | :--- | | **model** | `self.config.model` | `self.config.model` | | **temperature / top_p** | Excluded | Included (`self.config.temperature`, `self.config.top_p`) | | **max_tokens** | `max_completion_tokens` | `max_tokens` | | **reasoning_effort** | Included if defined | Excluded | | **response_format** | Optional | Optional | Sources: [mem0/llms/azure_openai_structured.py:77-101](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/azure_openai_structured.py#L77-L101) Response parsing in `AzureOpenAIStructuredLLM` branches based on whether tools were supplied in the request. If tools are active, `_parse_response` extracts content from `response.choices[0].message.content` and iterates over any returned `tool_calls`, parsing function arguments safely using `extract_json` and `json.loads`. Without tools, it returns the raw message content string directly. Sources: [mem0/llms/azure_openai_structured.py:125-153](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/azure_openai_structured.py#L125-L153) ## OpenAI and drop-in compatible providers ### OpenAI API Client and Drop-In Integrations The OpenAI adapter (`OpenAILLM`) extends `LLMBase` to manage completions and function calling through the standard OpenAI Python client. If `OPENROUTER_API_KEY` is present in the environment, it routes requests through OpenRouter using custom base URLs and headers (`HTTP-Referer`, `X-Title`). Otherwise, it resolves credentials via `self.config.api_key` or `OPENAI_API_KEY`, defaulting to `"gpt-5-mini"` when no model is specified. Sources: [mem0/llms/openai.py:14-54](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/openai.py#L14-L54) > [!WARNING] > OpenAI-compatible backends reject unrecognized parameters. `OpenAILLM` ensures that the `store` field is only transmitted when explicitly configured by the user, preventing rejections from third-party endpoints. Sources: [mem0/llms/openai.py:129-135](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/openai.py#L129-L135) Compatible providers implement identical initialization and parsing patterns while targeting custom endpoints. `DeepSeekLLM` defaults to the `"deepseek-chat"` model, resolving its API key and base URL from `DEEPSEEK_API_KEY` and `DEEPSEEK_API_BASE` (falling back to `https://api.deepseek.com`). `XAILLM` defaults to `"grok-4.3"`, utilizing `XAI_API_KEY` and `XAI_API_BASE` with a default URL of `https://api.x.ai/v1`. `VllmLLM` defaults to `Qwen/Qwen2.5-32B-Instruct` and requires explicit `vllm_base_url` configuration with a fallback API key of `"vllm-api-key"`. `LMStudioLLM` defaults to a local GGUF model path and uses `"lm-studio"` as its default API key. Sources: [mem0/llms/deepseek.py:13-42](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/deepseek.py#L13-L42), [mem0/llms/xai.py:13-42](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/xai.py#L13-L42), [mem0/llms/vllm.py:13-40](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/vllm.py#L13-L40), [mem0/llms/lmstudio.py:12-42](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/lmstudio.py#L12-L42) ```python if self.config.lmstudio_response_format: params["response_format"] = self.config.lmstudio_response_format elif response_format: params["response_format"] = response_format else: params["response_format"] = {"type": "json_object"} ``` Sources: [mem0/llms/lmstudio.py:102-108](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/lmstudio.py#L102-L108) ### Response Parsing and Tool Execution All adapters in this family share a common `_parse_response` implementation. When tools are provided in the generation request, the method constructs a dictionary containing the message content and an array of executed `tool_calls`, where function arguments are parsed from JSON strings using `extract_json`. When no tools are requested, it returns the raw response content string directly. Sources: [mem0/llms/deepseek.py:43-71](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/deepseek.py#L43-L71), [mem0/llms/lmstudio.py:43-71](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/lmstudio.py#L43-L71), [mem0/llms/xai.py:43-71](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/xai.py#L43-L71), [mem0/llms/openai.py:55-83](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/openai.py#L55-L83), [mem0/llms/vllm.py:42-70](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/vllm.py#L42-L70) | Provider Class | Default Model | Default Base URL / Environment Variable | | :--- | :--- | :--- | | **OpenAILLM** | `gpt-5-mini` | `https://api.openai.com/v1` (`OPENAI_BASE_URL`) | | **DeepSeekLLM** | `deepseek-chat` | `https://api.deepseek.com` (`DEEPSEEK_API_BASE`) | | **XAILLM** | `grok-4.3` | `https://api.x.ai/v1` (`XAI_API_BASE`) | | **VllmLLM** | `Qwen/Qwen2.5-32B-Instruct` | `vllm_base_url` config parameter | | **LMStudioLLM** | `lmstudio-community/Meta-Llama-3.1-70B-Instruct-GGUF/...` | `lmstudio_base_url` config parameter | Sources: [mem0/llms/openai.py:39-51](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/openai.py#L39-L51), [mem0/llms/deepseek.py:36-40](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/deepseek.py#L36-L40), [mem0/llms/xai.py:36-40](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/xai.py#L36-L40), [mem0/llms/vllm.py:36-39](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/vllm.py#L36-L39), [mem0/llms/lmstudio.py:35-41](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/lmstudio.py#L35-L41) ## Cloud provider abstractions ### AWS Bedrock Multi-Model Input Formatting and Parameter Mappings AWS Bedrock supports diverse model families through `BedrockLLM`, which normalizes generation requests and translates configuration parameters based on the target provider. Provider parameter mapping dictionaries explicitly map generic configuration keys to model-specific keys: | Provider Identifier | Parameter Mapping | | :--- | :--- | | `meta` | `max_tokens` $\rightarrow$ `max_gen_len` | | `ai21` | `max_tokens` $\rightarrow$ `maxTokens`, `top_p` $\rightarrow$ `topP` | | `mistral` | `max_tokens` $\rightarrow$ `max_tokens` | | `cohere` | `max_tokens` $\rightarrow$ `max_tokens`, `top_p` $\rightarrow$ `p` | | `amazon` | `max_tokens` $\rightarrow$ `maxTokenCount`, `top_p` $\rightarrow$ `topP` | | `anthropic` | `max_tokens` $\rightarrow$ `max_tokens`, `top_p` $\rightarrow$ `top_p` | Sources: [mem0/llms/aws_bedrock.py:261-276](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/aws_bedrock.py#L261-L276) Depending on the specific AWS model family selected, `_prepare_input` constructs distinct body structures: - **Amazon Nova Models:** Uses the Converse API format containing a `messages` array, `max_tokens` (defaulting to 5000), `temperature` (defaulting to 0.1), and optional nucleus sampling (`top_p`). - **Legacy Amazon Models:** Populates `inputText` alongside a `textGenerationConfig` object containing `maxTokenCount` and `temperature`. - **Anthropic Models:** Formats messages into a structured block with `anthropic_version` set to `"bedrock-2023-05-31"`, `max_tokens` defaulting to 2000, and `temperature` to 0.1. - **Cohere Command Models:** Replaces the standard `prompt` key with `message` when `"cohere.command"` is present in the model identifier. Sources: [mem0/llms/aws_bedrock.py:277-324](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/aws_bedrock.py#L277-L324) > [!NOTE] > When converting standard function tools for AWS Bedrock via `_convert_tool_format`, the engine extracts required parameters, constructs an `inputSchema` wrapping a JSON object type with properties, and stores the resulting tool specification under the `toolSpec` key. Sources: [mem0/llms/aws_bedrock.py:326-361](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/aws_bedrock.py#L326-L361) ### Gemini Message Reformatting and Tool Integration `GeminiLLM` instantiates clients using either the standard Google GenAI client or Vertex AI (`vertexai=True`), defaulting to `"gemini-2.0-flash"` when no model name is supplied. Sources: [mem0/llms/gemini.py:36-44](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/gemini.py#L36-L44) The adapter separates incoming chat history via `_reformat_messages`, isolating system prompts into `system_instruction` while converting remaining entries into `types.Content` structures with part lists and role assignments. Sources: [mem0/llms/gemini.py:92-115](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/gemini.py#L92-L115) ```python def remove_additional_properties(data): """Recursively removes 'additionalProperties' from nested dictionaries.""" if isinstance(data, dict): filtered_dict = { key: remove_additional_properties(value) for key, value in data.items() if not (key == "additionalProperties") } return filtered_dict else: return data ``` Sources: [mem0/llms/gemini.py:128-139](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/gemini.py#L128-L139) Tool definitions undergo recursive purging of `additionalProperties` constraints before being wrapped into `types.FunctionDeclaration` arrays and encapsulated within `types.Tool` objects. Sources: [mem0/llms/gemini.py:117-156](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/gemini.py#L117-L156) Tool choice configurations map API parameters directly to Gemini execution modes using `types.FunctionCallingConfigMode` options: - `auto` maps to `types.FunctionCallingConfigMode.AUTO` - `any` maps to `types.FunctionCallingConfigMode.ANY`, restricting execution via `allowed_function_names` - Any other value defaults to `types.FunctionCallingConfigMode.NONE` Sources: [mem0/llms/gemini.py:204-221](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/gemini.py#L204-L221) ## Local and universal runtime bridges ### Local Deployment via Ollama `OllamaLLM` integrates local models by wrapping the `ollama` Python library and accepting configuration variants including `BaseLlmConfig`, `OllamaConfig`, or raw dictionaries, defaulting to `"llama3.1:70b"` when model selection is omitted. Sources: [mem0/llms/ollama.py:16-42](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/ollama.py#L16-L42) ```python # Build parameters for Ollama params = { "model": self.config.model, "messages": messages, } # Handle JSON response format by using Ollama's native format parameter if response_format and response_format.get("type") == "json_object": params["format"] = "json" messages = [dict(m) for m in messages] if messages and messages[-1]["role"] == "user": messages[-1]["content"] += "\n\nPlease respond with valid JSON only." else: messages.append({"role": "user", "content": "Please respond with valid JSON only."}) params["messages"] = messages # Add options for Ollama (temperature, num_predict, top_p) options = { "temperature": self.config.temperature, "num_predict": self.config.max_tokens, "top_p": self.config.top_p, } params["options"] = options # Remove OpenAI-specific parameters that Ollama doesn't support params.pop("max_tokens", None) # Ollama uses different parameter names if tools: params["tools"] = tools response = self.client.chat(**params) return self._parse_response(response, tools) ``` Sources: [mem0/llms/ollama.py:113-144](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/ollama.py#L113-L144) > [!WARNING] > Ollama does not support OpenAI's `max_tokens` argument name. `OllamaLLM` explicitly pops `max_tokens` from the parameter dictionary and maps configuration limits to `num_predict` inside the nested `options` block. Sources: [mem0/llms/ollama.py:129-138](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/ollama.py#L129-L138) ### Universal Proxying via LiteLLM `LiteLLM` routes chat completion requests across diverse provider APIs through the `litellm` proxy package, defaulting to `"gpt-5-mini"` if no model string is specified. Sources: [mem0/llms/litellm.py:16-22](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/litellm.py#L16-L22) ```python if tools and not litellm.supports_function_calling(self.config.model): raise ValueError(f"Model '{self.config.model}' in litellm does not support function calling.") params = { "model": self.config.model, "messages": messages, "temperature": self.config.temperature, "top_p": self.config.top_p, } if self._uses_max_completion_tokens(self.config.model): params["max_completion_tokens"] = self.config.max_tokens else: params["max_tokens"] = self.config.max_tokens if response_format: params["response_format"] = response_format if tools: # TODO: Remove tools if no issues found with new memory addition logic params["tools"] = tools params["tool_choice"] = tool_choice response = litellm.completion(**params) ``` Sources: [mem0/llms/litellm.py:72-91](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/litellm.py#L72-L91) > [!IMPORTANT] > Before executing generation calls with tools enabled, `LiteLLM` verifies function support via `litellm.supports_function_calling(self.config.model)` and raises a `ValueError` if the target model lacks function-calling capabilities. Sources: [mem0/llms/litellm.py:72-73](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/litellm.py#L72-L73) ### LangChain Model Wrapping `LangchainLLM` wraps an instantiated LangChain chat model (`BaseChatModel`), requiring the `model` attribute to be supplied directly within the configuration object. Sources: [mem0/llms/langchain.py:13-24](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/langchain.py#L13-L24) ```python # Convert the messages to LangChain's tuple format langchain_messages = [] for message in messages: role = message["role"] content = message["content"] if role == "system": langchain_messages.append(("system", content)) elif role == "user": langchain_messages.append(("human", content)) elif role == "assistant": langchain_messages.append(("ai", content)) if not langchain_messages: raise ValueError("No valid messages found in the messages list") langchain_model = self.langchain_model if tools: langchain_model = langchain_model.bind_tools(tools=tools, tool_choice=tool_choice) response: AIMessage = langchain_model.invoke(langchain_messages, **kwargs) ``` Sources: [mem0/llms/langchain.py:76-96](https://github.com/blade47/mem0/blob/b7ad69afda6b6ed030347c66d48a13e4de9dec08/mem0/llms/langchain.py#L76-L96) ## Related - [Memory Lifecycle](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/python-core/memory-lifecycle) - [TypeScript LLM Adapters](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/technical/llm-providers/typescript-llm-adapters) ## Sitemap See the full [sitemap](https://www.doc0.dev/docs/faa36707-7c28-4f69-a18f-700ff61c704e/llms.txt) for all pages in this wiki.