Fix examples broken by SDK 5.12 and platform changes; add gap-filling samples - #81
Open
abhishekmishragithub wants to merge 1 commit into
Open
Fix examples broken by SDK 5.12 and platform changes; add gap-filling samples#81abhishekmishragithub wants to merge 1 commit into
abhishekmishragithub wants to merge 1 commit into
Conversation
… samples Fixes: - governance: the sample imported a smallestai.agentic module that has never existed and used invented node/tool/app signatures. Rewritten on the real crew surface (OutputCrewNode subclass, ToolRegistry.discover, AtomsCrewApp setup_handler) with working TealTiger guardrails: PII is redacted from the caller turn before the LLM sees it and from the reply before TTS, with a per-call turn cap. - Mobile examples (react-native x2, iOS): the v1 agent-versioning flow (drafts/versions/activate) now 404s. Migrated to the v2 branches flow: resolve the live branch, PUT the branch draft, publish, and poll the branch until the async security scan closes the draft. Raw REST responses use camelCase and _id, handled in the py/ts/swift clients. - Hydra: the voice catalog changed and the old ids are rejected. Presets, voice picker, and docs move to current voices (aria, maya, sterling, ...). - voice-cloning and sdk-usage READMEs: replaced Coming Soon stubs and removed-method snippets (add_voice, WavesClient.synthesize) with the current create_voice_clone / list_voice_clones / synthesize_tts patterns and doc links. New examples for documented-but-uncovered features: - text-to-speech/word-timestamps-live: per-word timing frames on the TTS WebSocket, printed as live captions. - llm/tool-calling: Electron function calling through the OpenAI-compatible /waves/v1 endpoint, plus an llm section index. - speech-to-speech/python-quickstart: minimal headless Hydra client with real-time pacing and VAD turn close. - voice-agents/web_call_session: the web-call session API (token, room, conversation id) with the calls.get follow-up.
|
crim doesn't review pull requests automatically here. Comment |
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes (broken today)
smallestai.agenticmodule that has never existed in any SDK release, and used invented signatures throughout (generate_response(transcript),AtomsCrewApp(agent=...), voice/language on the node). Rewritten on the real crew surface with working TealTiger guardrails: PII is redacted from the caller turn before the LLM sees it and from the reply before TTS, plus a per-call turn cap.atomsRest.ts, iOSAtomsRest.swift): the v1 versioning flow (/drafts,/versions/{v}/activate) now returns 404. Migrated to the v2 branches flow: resolve the live branch,PUT .../branches/{id}/draft, publish, then poll the branch until the async security scan closes the draft. Raw REST responses use camelCase and_id; the clients handle both key styles.text-to-speech/voice-cloningandtext-to-speech/sdk-usagereplaced their "Coming Soon" text and removed-method snippets (add_voice,WavesClient.synthesize) with the currentcreate_voice_clone/list_voice_clones/synthesize_ttspatterns.New examples (documented features with no cookbook coverage)
text-to-speech/word-timestamps-live: per-word timing frames on the TTS WebSocket, printed as live captions while the audio saves to WAV.llm/tool-calling(newllm/section): Electron function calling through the OpenAI-compatible/waves/v1endpoint.speech-to-speech/python-quickstart: minimal headless Hydra client (session.configure, real-time paced audio, VAD turn close, transcript streaming).voice-agents/web_call_session: the new web-call session API (client.atoms.web_call.*) returning the room token/host and thecalls.getfollow-up.How it was tested
All Python paths ran against api.smallest.ai with a real org key on smallestai 5.12.0: the governance agent answered a turn with the SSN in the transcript redacted before the LLM call (verified in logs);
setup_agent.pycreated an agent, edited and published a branch draft through the scan poll, and went live (test agent archived afterward); word-timestamps returned 12 timed words plus audio; Electron executed a real tool round trip; web_call minted a live session; the Hydra quickstart streams and prints the reply transcript. Swift and TS clients are parse-checked; mobile flows still need a device test.Known platform issues found (escalated separately, not fixable here)
wss /atoms/v1/agent/connectrejects every documented auth variant with 401 (register-call token and api-key+agent_id), which affects the merged mobile examples and PR recipe(voice-agents): agent_ws_client_tools — Atoms Agent WS client-tools sample #78;atoms.web_callLiveKit sessions work, suggesting the realtime transport moved.response.output_audio.deltaframes (July runs returned audio on the same flow).voice-agents/atoms_sdk_web_agentuses npmatoms-client-sdkwhile docs now describe@smallest-ai/agent-sdk; both exist, migration is a follow-up.