UN-3987 [FIX] Send LLMWhisperer V2 params under the correct names - #2236
Merged
chandrasekharan-zipstack merged 8 commits intoAug 18, 2026
Merged
Conversation
chandrasekharan-zipstack
force-pushed
the
UN-3987-fix-line-splitter-strategy-key
branch
from
August 11, 2026 08:09
0b28344 to
151b688
Compare
The adapter read the line splitter strategy under `line_spitter_strategy` while its JSON schema stores it as `line_splitter_strategy`, so the user's choice never applied. The same misspelling was also passed to the client, which forwarded a query param the service does not read. The page separator keeps its misspelled config key since existing adapter configs are stored under it, but is now sent under the client's corrected kwarg. Both need llmwhisperer-client 2.8.0 or newer. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Dra3Xevzb5oYtMz9fhj8iG
Without it, usage reports record the service default for every Unstract-originated extraction, leaving no way to cross reference a row back to a document. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Dra3Xevzb5oYtMz9fhj8iG
chandrasekharan-zipstack
force-pushed
the
UN-3987-fix-line-splitter-strategy-key
branch
from
August 11, 2026 08:39
151b688 to
8b74d6f
Compare
chandrasekharan-zipstack
marked this pull request as ready for review
August 12, 2026 11:46
![uv-lock-automation[bot]](https://github.githubassets.com/images/gravatars/gravatar-user-420.png){kind=avatar}
|
Contributor
|
| Filename | Overview |
|---|---|
| unstract/sdk1/src/unstract/sdk1/adapters/x2text/llm_whisperer_v2/src/helper.py | Correctly maps and validates the strategy, translates the legacy page-separator key, and adds the input basename. |
| unstract/sdk1/src/unstract/sdk1/adapters/x2text/llm_whisperer_v2/src/constants.py | Separates persisted configuration names from corrected client parameter names and enumerates accepted strategies. |
| unstract/sdk1/tests/test_llm_whisperer_v2_params.py | Covers supported and invalid strategy handling plus legacy page-separator translation. |
| unstract/sdk1/pyproject.toml | Requires llmwhisperer-client 2.8.1, matching the version resolved by every affected lockfile. |
| docker/scripts/uv-lock-gen/uv-lock.sh | Fixes the previously reported transitive-lock gap by inspecting local sources in the complete resolved lock graph and covering all lockfile projects. |
| unstract/sdk1/uv.lock | Now consistently resolves llmwhisperer-client 2.8.1 under the updated SDK requirement. |
Flowchart
%%{init: {'theme': 'neutral'}}%%
flowchart LR
C[Stored adapter config] --> H[LLMWhisperer V2 helper]
H --> V[Validate line splitter strategy]
V --> P[Correct client parameter names]
P --> L[llmwhisperer-client 2.8.1]
L --> S[LLMWhisperer service]
D[Changed local pyproject] --> G[Lock generator reads pyproject and uv.lock graph]
G --> R[Regenerate every affected project lock]
Reviews (5): Last reviewed commit: "UN-3987 [FIX] Read the whole local depen..." | Re-trigger Greptile
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
…mers The uv-lock bot only checks a directory's own pyproject and its immediate local path dependencies, so an sdk1-only dependency bump left five lockfiles pinning the pre-2.8.0 client with a stale `>=2.6.2` specifier. `uv sync --locked` rejected those directories. Adds sdk1, tool-registry and workflow-execution to the bot's directory list so a future sdk1 bump relocks them without a manual pass. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EVoa8BMCNnhdkATd7A9U4J
…service accepts The setting has never reached the service, so stored values were never validated and the JSON schema accepted free text. Now that the value is actually sent, anything outside left-priority / mid-priority / right-priority makes the service reject the extraction with a 400. Adds the enum to the schema so the field renders as a dropdown, and falls back to the default with a warning when an existing config holds an unsupported value. The fallback preserves what those configs resolve to today, and the warning surfaces how many are affected without decrypting stored metadata. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EVoa8BMCNnhdkATd7A9U4J
2.8.1 sends the page separator under both the corrected and the legacy query key, so the adapter no longer depends on the LLMWhisperer service being new enough to read the corrected one. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EVoa8BMCNnhdkATd7A9U4J
johnyrahul
approved these changes
Aug 18, 2026
has_dependency_changes only walked one level of [tool.uv.sources], so a change to unstract/sdk1 relocked unstract/filesystem (a direct dependant) but not unstract/connectors, which reaches sdk1 through it. That is why the root and connectors locks were already stale at llmwhisperer-client 2.7.0 before this branch existed. Read the resolved graph out of each uv.lock instead: uv has already computed the closure there, so no recursion is needed. The pyproject is still read so a dependency added but not yet locked is not missed. Against this branch's sdk1 change, the old walk skips '.' and unstract/connectors; the new one selects both. Also cover unstract/tool-sandbox and tool-sidecar, the two remaining directories with a uv.lock that the bot never visited. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01EVoa8BMCNnhdkATd7A9U4J
|
Contributor
Unstract test resultsPer-group results
Critical paths
|
chandrasekharan-zipstack
deleted the
UN-3987-fix-line-splitter-strategy-key
branch
August 18, 2026 15:40
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
2 participants
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.
What
The LLMWhisperer V2 adapter's line splitter strategy setting has never been applied. Two breaks stacked:
line_splitter_strategy, butWhispererConfig.LINE_SPLITTER_STRATEGYwas"line_spitter_strategy", soconfig.get()never matched and theleft-prioritydefault was always used.whisper()kwarg, so the client sent query paramline_spitter_strategy, which the service does not read and Flask drops silently.Fixing either alone changes nothing.
Changes
LINE_SPLITTER_STRATEGYis now"line_splitter_strategy"— matches the schema key and the wire param.WhispererConfiggainsPAGE_SEPARATOR_PARAMandFILE_NAMEfor the client kwarg names.page_seperatorstays as the config key (existing adapter configs in the DB are stored under it, no migration needed) but is now sent aspage_separator.file_nameis now sent, set to the basename of the file being extracted. Previously the adapter never sent it, so every Unstract-originated row in LLMWhisperer usage reports showed the service default. Separate commit if you'd rather drop it.llmwhisperer-clientpinned to>=2.8.0.Blocked on the client release
>=2.8.0does not exist yet — it is the release that carries Zipstack/llm-whisperer-python-client#34, which adds the correctly-namedline_splitter_strategy,page_separatorandfile_namekwargs. On an older client these kwargs raiseTypeError, so this must not merge before that release. Draft until then. On un-drafting: confirm the actual version number and regenerate the lockfiles.Ordering: unstract-llm-whisperer#721 (service accepts
page_separator) → llm-whisperer-python-client#34 release → this PR.Behaviour change to expect
The service validates the strategy and returns
400for anything outsideleft-priority/right-priority/mid-priority. Previously an out-of-range value was silently ignored, so a bad stored value becomes a hard error once the value actually arrives.Users who never touched the setting are unaffected — adapter, client and service all default to
left-priority.Verification
tests/test_llm_whisperer_v2_params.py— 2 tests, both pass: the strategy is read from the schema key, and the page separator is read from the legacy config key but emitted under the corrected kwarg.Not verified end to end: the extraction-output change needs the client release and a run against a deployed service.
Jira: UN-3987, LW-406
🤖 Generated with Claude Code
https://claude.ai/code/session_01Dra3Xevzb5oYtMz9fhj8iG