chore(deps): deepgram-sdk 6.0.1 -> 7.6.0#12
Merged
Conversation
Separate from the general dependency refresh on purpose: this is the axis most
likely to break, so it gets to fail on its own.
The app's v1 surface is unaffected. Comparing connect() signatures across the
major, NOTHING was removed and one param was added (`diarize_model`), so every
call site keeps working. `_SDK_CONNECT_KWARGS` is computed from the installed SDK
rather than hardcoded, so the named/passthrough split re-derived itself with no
edit — which is the whole reason it was written that way.
The only breakage was in this repo's own test helper:
TypeError: DeepgramClientEnvironment.__init__() missing 1 required
keyword-only argument: 'agent_rest'
7.0 added an endpoint field, and the helper named its fields positionally. It now
enumerates the constructor signature instead, so the next regeneration that adds
a URL field costs nothing. Worth noting the wire-contract tests added last week
are what caught this at all; the rest of the suite passed.
Also unblocks the Flux work. 7.x ships `listen.v2.connect()` with the full Flux
parameter surface — eot_threshold, eager_eot_threshold, eot_timeout_ms,
language_hint, keyterm, numerals, profanity_filter — so Flux does not need the
parallel raw-WebSocket path that would have been necessary on 6.x.
Verified:
- 113 passed, 1 skipped.
- App boots on 7.6.0; GET / returns 200; UI probe passes.
- Real WebSocket handshake still carries keyterm/passthrough params correctly.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Separate from #11 on purpose: this is the axis most likely to break, so it gets to fail on its own.
The app's v1 surface is unaffected
Comparing
listen.v1.connect()signatures across the major:Nothing was removed, so every call site keeps working.
_SDK_CONNECT_KWARGSis computed from the installed SDK rather than hardcoded, so the named/passthrough split re-derived itself with no edit — which is the whole reason it was written that way.The only breakage was our own test helper
7.0 added an endpoint field and the helper named its fields explicitly. It now enumerates the constructor signature, so the next regeneration that adds a URL field costs nothing.
Worth noting: the wire-contract tests added in #8 are what caught this. The rest of the suite passed clean.
This unblocks Flux
7.x ships
listen.v2.connect()with the full Flux parameter surface:So Flux STT can go through the SDK rather than needing the parallel raw-WebSocket path that 6.x would have forced. 7.6.0 specifically adds Flux
numerals.Verification
113 passed, 1 skipped.GET /returns 200; UI probe passes in batch mode.keytermand the passthrough params correctly.Live verification against prod to follow the deploy — the SDK change touches the real wire path, so unit tests are not sufficient proof here.