> ## Documentation Index
> Fetch the complete documentation index at: https://docs.inputapi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Example index and learning paths

> Find APIs, complete scripts, and validation boundaries by integration goal.

Start with the [quickstart](/en/quickstart) to verify the address, credentials, and model, then select one complete example. Sources live in the documentation directory's `examples/`; copy environment variables from the relevant guide. Run `npm ci` there for Node examples. See the [Python guide](/en/guides/python) for Python setup and the documentation directory’s `README.md` for complete local checks.

| Goal | Guide / runnable entry point | Key acceptance check |
| - | - | - |
| Minimal text | [Python](/en/guides/python) / [TypeScript](/en/guides/typescript) | Base URL, model, complete output |
| Framework integration | [Vercel AI SDK](/en/guides/ai-sdk) | Actual protocol path, SDK retries, completion state |
| Conversation memory | [Multi-turn conversations](/en/guides/conversations) | Complete history within one protocol |
| Client tools | [Tool calling](/en/guides/tool-calling) / [streaming tools](/en/guides/streaming-tools) | Call IDs, arguments, result association, round limits |
| JSON results | [Structured output](/en/guides/structured-output) | Schema validation and incomplete outcomes |
| Document retrieval | [RAG](/en/guides/rag) | Embeddings, Rerank, source attribution |
| Images and audio | [Vision](/en/guides/vision) / [image generation](/en/guides/image-generation) / [speech](/en/guides/speech) / [transcription](/en/guides/speech-to-text) | Model capability, file types, complete downloads |
| Video and async tasks | [Video](/en/guides/video-generation) / [tasks](/en/guides/tasks) | Existing task IDs, terminal states, polling limits |
| Persistent connections | [Responses WebSocket](/en/guides/responses-websocket) / [Realtime](/en/guides/realtime) | Handshake, event correlation, audio format, closure |
| Native protocols | [Claude SSE](/en/guides/native-claude-streaming) / [Gemini SSE](/en/guides/native-gemini-streaming) | Native blocks, signatures, terminal events |
| Operational diagnostics | [Rate limits](/en/guides/rate-limits) / [observability](/en/guides/observability) | Read-only probes, bounded retries, sanitized logs |

## From examples to production

1. Run documentation validation against local fixtures to check code and page consistency.
2. Use isolated test settings for one minimal target-deployment request, then verify each required extension separately.
3. Test business failures, timeouts, streaming interruption, and duplicate submission handling before setting queue, concurrency, and logging policies.
4. Record target-service results using the [production checklist](/en/guides/production). Local passes do not replace real model verification.

Fields are centralized in the [API reference](/en/api-reference/overview). Check [coverage](/en/guides/api-coverage) before selecting an endpoint. These scripts work independently of Mintlify's runtime; retain the same examples and OpenAPI file when migrating documentation platforms.

[阅读中文](/zh/guides/recipes)

[Prompt caching](/en/guides/prompt-caching) · [Reasoning models](/en/guides/reasoning-models) · [Context management](/en/guides/context-management)

## Files, search, and generation parameters

| Task | Guide | Runnable script |
| - | - | - |
| Inline PDF summary | [File inputs](/en/guides/file-inputs) | `examples/file-input.mjs` |
| Native web search with URL citations | [Search](/en/guides/web-search) | `examples/web-search.mjs` |
| Inspect token log probabilities | [Generation parameters](/en/guides/generation-parameters) | `examples/logprobs.mjs` |

Confirm model and native-channel support before running; also see [data controls](/en/guides/data-controls).

## Audio/video input and text revisions

| Goal | Guide | Script |
| - | - | - |
| Ask about a recording | [Audio input](/en/guides/audio-input) | `examples/audio-input.mjs` |
| Analyze an existing video | [Video understanding](/en/guides/video-understanding) | `examples/video-input.mjs` |
| Hint at retained source during revisions | [Predicted outputs](/en/guides/predicted-outputs) | `examples/predicted-output.mjs` |

Each example makes one request. Verify real results with the target model; see [troubleshooting](/en/guides/troubleshooting).

## Tool policies and JSON validation

* [Tool selection and parallel controls](/en/guides/tool-controls): `examples/tool-policy.mjs`, four policies, at most two turns and one read-only call.
* [JSON mode](/en/guides/json-mode): `examples/json-mode.mjs`, Chat / Responses protocols with separate completion and business-field validation.

## Native Gemini built-in tools

[Three-mode example](/en/guides/gemini-builtins): `examples/gemini-builtins.mjs` selects search, URL context, or upstream code execution. One request per run; checks task evidence as well as completion.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.