Skip to content

docs: add transcript readers for unknown speakers and timestamps - #64

Merged
JJasonSun merged 1 commit into
mainfrom
docs/transcript-reading
Sep 15, 2026
Merged

JJasonSun merged 1 commit into
mainfrom
docs/transcript-reading

Conversation

@JJasonSun

Copy link
Copy Markdown
Collaborator

The Calls guide links to runnable Python/HTTP and TypeScript SDK transcript readers. They iterate recipient attempts, label the original speaker, display null offsets as time unavailable, preserve zero offsets, and leave input data unchanged. Unknown speech is not treated as evidence that the recipient answered.

Refs CALLE-AI/awesome-phone-call-agents#597.

Validation:

  • Both standalone synthetic checks pass for bot, user, unknown, zero/null offsets, and an empty transcript. They make no network requests and are included in pnpm run validate.
  • Read the same completed real call using published Python and TypeScript SDK 0.7.0; both readers produced identical output for all three returned turns. This made two read requests and no new call. The live sample does not establish unknown-speaker or null-offset production behavior; those cases are covered by synthetic checks against the documented contract.
  • Python 3.11.16 and Node 26.7.0 execution passed; strict TypeScript checking passed.
  • Typecheck, build, distribution verification, 13 deployment tests, and 22 Chromium tests passed. Generated Calls Markdown includes the reader links and guidance.

No API or SDK behavior changes and no package release required. Production guide verification remains pending merge and deployment.

@JJasonSun
JJasonSun merged commit 43337c2 into main Sep 15, 2026
1 check passed
@JJasonSun
JJasonSun deleted the docs/transcript-reading branch September 15, 2026 10:04
@JJasonSun

Copy link
Copy Markdown
Collaborator Author

Completed in #64, merged as 43337c2. Production deployment succeeded: https://github.com/CALLE-AI/calle-docs/actions/runs/34955985060.

Live verification confirms the Calls transcript-reading section, both runnable reader links, calls.md, and llms-full.txt. Both public main example files match the tested source and their synthetic checks pass. Python and TypeScript SDK 0.7.0 real-result reads produced identical output for all three returned turns; no new call was created. Synthetic checks separately cover unknown speakers and null offsets without claiming those conditions were observed in the real sample.

The readers preserve speaker labels and distinguish missing timestamps from zero. All acceptance criteria are complete; no API behavior or SDK release change is involved.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant