docs: add FunASR voice input example

[LauraGPT]

Summary

• document a small voice-input pattern for CrewAI: transcribe a recording with an OpenAI-compatible STT endpoint before crew.kickoff
• use a local FunASR/SenseVoice server as the concrete example
• keep the transcript as normal text inputs so no new CrewAI runtime dependency or ASR provider abstraction is required

Related to #5983. This PR covers the voice-input/transcription slice only; it does not claim to implement TTS or the full voice-agent workflow.

Validation

• git diff --check
• content assertions for the new section, FunASR command, OpenAI transcription call, and crew.kickoff example

mintlify broken-links was not run because the Mintlify CLI is not installed in this remote checkout; this PR adds no new links.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: d0c18efa-b5ff-4edf-9ac0-aea2a55c8648

📥 Commits

Reviewing files that changed from the base of the PR and between 65156a0 and d2bc103.

📒 Files selected for processing (1)

• docs/edge/en/concepts/files.mdx

🚧 Files skipped from review as they are similar to previous changes (1)

• docs/edge/en/concepts/files.mdx

---

📝 Walkthrough

Walkthrough

Adds a “Transcribing Audio Before Kickoff” section describing a local FunASR server, OpenAI-compatible audio transcription, and passing the resulting text into crew.kickoff(inputs=...).

Changes

Audio transcription documentation

Layer / File(s)	Summary
Pre-kickoff transcription workflow docs/edge/en/concepts/files.mdx	Documents starting a local FunASR server, transcribing an audio file through an OpenAI-compatible client, and supplying transcription.text to crew inputs.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately summarizes the main change: adding a FunASR voice-input example in the docs.
Description check	✅ Passed	The description matches the documented voice-input transcription workflow and the stated scope of the PR.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[coderabbitai]

Actionable comments posted: 1

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@docs/edge/en/concepts/files.mdx`:
- Around line 108-112: Update the FunASR startup command in the local setup
example to bind explicitly to 127.0.0.1, preserving the existing model, device,
and port options.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: d04c690c-a2cb-41bc-a972-2516fdd5948f

📥 Commits

Reviewing files that changed from the base of the PR and between 69c0308 and e9babcd.

📒 Files selected for processing (1)

• docs/edge/en/concepts/files.mdx

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

docs/edge/en/concepts/files.mdx (1)

111-121: 🎯 Functional Correctness | 🟡 Minor | ⚡ Quick win

Use the same loopback address in the client and server examples.

The server binds to IPv4 127.0.0.1, but the client connects to localhost. On systems where localhost resolves to IPv6 ::1 first, the request can fail because the server is not listening on IPv6. Use http://127.0.0.1:8000/v1 or bind the server to both address families.

Proposed fix

 client = OpenAI(
     api_key="EMPTY",
-    base_url="http://localhost:8000/v1",
+    base_url="http://127.0.0.1:8000/v1",
 )

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@docs/edge/en/concepts/files.mdx` around lines 111 - 121, Update the OpenAI
client’s base_url in the transcription example to use the same IPv4 loopback
address as the FunASR server command: 127.0.0.1 on port 8000 with the /v1 path,
replacing localhost.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Outside diff comments:
In `@docs/edge/en/concepts/files.mdx`:
- Around line 111-121: Update the OpenAI client’s base_url in the transcription
example to use the same IPv4 loopback address as the FunASR server command:
127.0.0.1 on port 8000 with the /v1 path, replacing localhost.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: fee20397-856e-4969-a142-4ced86a3abb8

📥 Commits

Reviewing files that changed from the base of the PR and between e9babcd and f30bf50.

📒 Files selected for processing (1)

• docs/edge/en/concepts/files.mdx

[LauraGPT]

Rebased this docs-only PR onto current main to clear the behind state. Re-ran focused validation on the refreshed head:

• git diff --check origin/main...HEAD
• content assertions for the FunASR section, IPv4 loopback server command, OpenAI-compatible transcription client, and crew.kickoff(inputs=...) usage

The PR still intentionally covers only the voice-input/transcription documentation slice and no longer claims to close the broader voice-agent/TTS issue.