> ## Documentation Index > Fetch the complete documentation index at: https://docs.fallom.com/llms.txt > Use this file to discover all available pages before exploring further. # API Reference > Complete SDK reference for Fallom ### Initialize ```python theme={null} fallom.init(api_key=None, base_url=None, capture_content=True) ``` Initialize the SDK. Call this **before** importing LLM libraries for auto-instrumentation. | Parameter | Type | Description | | ----------------- | ---- | ----------------------------------------------------------- | | `api_key` | str | Your Fallom API key (or use `FALLOM_API_KEY` env var) | | `base_url` | str | API endpoint (default:`https://spans.fallom.com`) | | `capture_content` | bool | Whether to capture prompt/completion text (default: `True`) | *** ### Get Model ```python theme={null} fallom.models.get(config_key, session_id, version=None, fallback=None) -> str ``` Get model assignment for a session. | Parameter | Type | Description | | ------------ | ---- | -------------------------------------------------- | | `config_key` | str | Your config name from the dashboard | | `session_id` | str | Unique session/conversation ID (sticky assignment) | | `version` | int | Pin to specific version (default: latest) | | `fallback` | str | Model to return if anything fails | *** ### Get Prompt ```python theme={null} fallom.prompts.get(prompt_key, variables=None, version=None) -> PromptResult ``` Get a managed prompt. | Parameter | Type | Description | | ------------ | ---- | -------------------------------------------------- | | `prompt_key` | str | Your prompt key from the dashboard | | `variables` | dict | Template variables (e.g., `{"user_name": "John"}`) | | `version` | int | Pin to specific version (default: latest) | **Returns:** `PromptResult` with `key`, `version`, `system`, `user` *** ### Get Prompt A/B Test ```python theme={null} fallom.prompts.get_ab(ab_test_key, session_id, variables=None) -> PromptResult ``` Get a prompt from an A/B test (sticky assignment). | Parameter | Type | Description | | ------------- | ---- | ------------------------------------------------------ | | `ab_test_key` | str | Your A/B test key from the dashboard | | `session_id` | str | Unique session/conversation ID (for sticky assignment) | | `variables` | dict | Template variables | **Returns:** `PromptResult` with `key`, `version`, `system`, `user`, `ab_test_key`, `variant_index` *** ### Set Session ```python theme={null} fallom.trace.set_session(config_key, session_id, customer_id=None) ``` Set trace context. All subsequent LLM calls will be tagged with this `config_key` and `session_id`. *** ### Clear Session ```python theme={null} fallom.trace.clear_session() ``` Clear trace context. *** ### Record Custom Metrics ```python theme={null} fallom.trace.span(data, config_key=None, session_id=None) ``` Record custom business metrics. | Parameter | Type | Description | | ------------ | ---- | -------------------------------------- | | `data` | dict | Metrics to record | | `config_key` | str | Optional if `set_session()` was called | | `session_id` | str | Optional if `set_session()` was called | *** ### Supported LLM Providers Auto-instrumentation available for: * OpenAI (+ OpenAI-compatible APIs: OpenRouter, LiteLLM, vLLM, Ollama, etc.) * Anthropic * Cohere * AWS Bedrock * Google Generative AI * Mistral AI * LangChain * Replicate * Vertex AI Install the corresponding `opentelemetry-instrumentation-*` package for your provider. You must use the official SDK for your provider. Raw HTTP requests (e.g., `requests.post()`) will not be traced. For OpenAI-compatible APIs, use the OpenAI SDK with a custom `base_url`. ### Initialize ```typescript theme={null} fallom.init({ apiKey, baseUrl, captureContent }) ``` Initialize the SDK. Call once at app startup. | Option | Type | Description | | ---------------- | ------- | ----------------------------------------------------------- | | `apiKey` | string | Your Fallom API key (or use `FALLOM_API_KEY` env var) | | `baseUrl` | string | API endpoint (default:`https://spans.fallom.com`) | | `captureContent` | boolean | Whether to capture prompt/completion text (default: `true`) | *** ### Create Session ```typescript theme={null} fallom.session(options): FallomSession ``` Create a session-scoped tracer. Returns a `FallomSession` instance. | Option | Type | Description | | ------------ | ------ | ----------------------------------------------- | | `configKey` | string | Your config name from the dashboard | | `sessionId` | string | Unique session/conversation ID | | `customerId` | string | Optional user identifier for per-user analytics | *** ### FallomSession Methods ```typescript theme={null} const session = fallom.session({ configKey, sessionId, customerId }); ``` | Method | Description | | ------------------------ | --------------------------------------------------- | | `wrapOpenAI(client)` | Wrap OpenAI client for automatic tracing | | `wrapAnthropic(client)` | Wrap Anthropic client for automatic tracing | | `wrapGoogleAI(model)` | Wrap Google AI model for automatic tracing | | `wrapAISDK(ai)` | Wrap Vercel AI SDK module | | `wrapMastraAgent(agent)` | Wrap Mastra agent for automatic tracing | | `traceModel(model)` | Wrap a Vercel AI SDK model directly (PostHog style) | | `getModel(options?)` | Get model assignment for A/B testing | | `getContext()` | Get the session context object | *** ### Wrap OpenAI ```typescript theme={null} session.wrapOpenAI(client) ``` Wrap OpenAI client for automatic tracing. Works with any OpenAI-compatible API. *** ### Wrap Anthropic ```typescript theme={null} session.wrapAnthropic(client) ``` Wrap Anthropic client for automatic tracing. *** ### Wrap Google AI ```typescript theme={null} session.wrapGoogleAI(model) ``` Wrap Google AI model for automatic tracing. *** ### Wrap Vercel AI SDK ```typescript theme={null} session.wrapAISDK(ai) ``` Wrap Vercel AI SDK module. Returns `{ generateText, streamText, generateObject, streamObject }`. *** ### Trace Model (PostHog Style) ```typescript theme={null} session.traceModel(model) ``` Wrap a Vercel AI SDK model directly for automatic tracing. Returns a traced model that can be used with the original AI SDK functions. ```typescript theme={null} const model = session.traceModel(createOpenAI()("gpt-4o")); await generateText({ model, prompt: "Hello!" }); ``` *** ### Get Model (Session-scoped) ```typescript theme={null} session.getModel(options?): Promise ``` Get model assignment for A/B testing using the session's config key. | Option | Type | Description | | ---------- | ------ | --------------------------------- | | `fallback` | string | Model to return if anything fails | | `version` | number | Pin to specific config version | *** ### Get Model (Standalone) ```typescript theme={null} fallom.models.get(configKey, sessionId, options?): Promise ``` Get model assignment for A/B testing without a session. | Parameter | Type | Description | | ------------------ | ------ | ----------------------------------- | | `configKey` | string | Your config name from the dashboard | | `sessionId` | string | Unique session/conversation ID | | `options.fallback` | string | Model to return if anything fails | *** ### Get Prompt ```typescript theme={null} fallom.prompts.get(promptKey, options?): Promise ``` Get a managed prompt. | Parameter | Type | Description | | ------------------- | ------ | ------------------------------------------------- | | `promptKey` | string | Your prompt key from the dashboard | | `options.variables` | object | Template variables (e.g., `{ userName: "John" }`) | | `options.version` | number | Pin to specific version (default: latest) | *** ### Get Prompt A/B Test ```typescript theme={null} fallom.prompts.getAB(abTestKey, sessionId, options?): Promise ``` Get a prompt from an A/B test. | Parameter | Type | Description | | ------------------- | ------ | ------------------------------------ | | `abTestKey` | string | Your A/B test key from the dashboard | | `sessionId` | string | Session ID for sticky assignment | | `options.variables` | object | Template variables |