> For the complete documentation index, see [llms.txt](https://robodex.gitbook.io/robodex-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://robodex.gitbook.io/robodex-docs/product-api/ingest.md).

# Ingest Pipeline

The ingest route lives in `apps/api/routes/ingest.py`.

It accepts a natural-language task prompt and tries to produce useful reference context for downstream robot design.

## `POST /ingest`

Main behavior:

1. create a job id,
2. log the raw prompt,
3. call `DemoService.analyze_prompt(req.prompt)`,
4. require at least one generated search query,
5. select a reference source,
6. optionally dispatch motion analysis,
7. persist the ingest job in `workspace_store`.

## Prompt analysis

The prompt analysis step extracts:

* task goal,
* constraints,
* search queries,
* possible motion-reference needs.

The design route later uses `plan.task_goal` or `selected_query` as the robot generation prompt.

## Reference selection

The route supports three practical states:

| State             | Meaning                                                                   |
| ----------------- | ------------------------------------------------------------------------- |
| YouTube selected  | A reference video was found and selected                                  |
| DROID fallback    | YouTube lookup failed or was unavailable, so a fallback reference is used |
| reference skipped | `SKIP_YOUTUBE_SEARCH=1`, so the route skips external search               |

## Motion analysis dispatch

If a YouTube video is selected, the route can try a GVHMR dispatch for human motion analysis. This is best-effort and should be treated as enrichment, not a hard dependency for body generation.

## Persistence

The persisted ingest job records:

* prompt,
* plan,
* selected query,
* source,
* selected video/reference,
* status,
* error details when applicable.

`POST /designs/generate` consumes the stored ingest job by id.

## Where ingest ends

Ingest does not generate robot bodies. It prepares task context. Robot body generation starts when the design route calls an agent loop.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://robodex.gitbook.io/robodex-docs/product-api/ingest.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.