Run IBM Watson transcription with Sapat in Daytona

[katlark]

/claim #13

Summary

• add a long-form Daytona guide for running Sapat with the new IBM Watson Speech to Text provider
• add a companion definition for synchronous speech recognition
• add a first-time author profile and original SVG workflow diagram
• link the companion Sapat provider implementation: Add IBM Watson Speech to Text provider nibzard/sapat#28

Companion implementation

• nibzard/sapat#28 adds --api ibm, posts converted audio to IBM Watson Speech to Text POST /v1/recognize, reads credentials from environment variables, and includes mocked unit coverage for request construction, response parsing, missing configuration, and CLI routing.

Validation

Daytona content PR:

• ./node_modules/.bin/markdownlint guides/20260520_ibm_watson_sapat_transcription_daytona.md definitions/20260520_definition_synchronous_speech_recognition.md authors/kat_lark.md
• python3 -c "import xml.etree.ElementTree as ET; ET.parse(\"guides/assets/20260520_ibm_watson_sapat_transcription_daytona_img1.svg\"); print(\"svg ok\")"
• git diff --check
• wc -w guides/20260520_ibm_watson_sapat_transcription_daytona.md definitions/20260520_definition_synchronous_speech_recognition.md authors/kat_lark.md (1,858 words total; guide is 1,625 words)

Companion Sapat PR:

• .venv/bin/python -m unittest discover -s tests -v
• .venv/bin/python -m py_compile src/sapat/script.py src/sapat/transcription/watson.py tests/test_watson_transcription.py tests/test_script.py
• .venv/bin/python -m sapat.script --help
• git diff --check

Notes

• npm run lint:md still fails on pre-existing markdownlint issues in older repository files outside this PR, including articles/20241212_Prebuilds_in_Daytona.md, definitions/20240819_definition_embedding.md, and guides/20240819_history_embedding.md.
• Sources checked: Sapat README/source, IBM Watson Speech to Text getting started docs, IBM synchronous HTTP docs, and IBM basic recognition request docs.
• No API keys, recordings, payment details, private account data, or local sensitive paths are included.

[katlark]

Closing this IBM Watson submission because another contributor registered and opened the same provider slice just before my PR (#223 / nibzard/sapat#27). I do not want to duplicate their work. I am pivoting to a non-overlapping Together AI transcription provider/workflow instead.