Repository navigation
Add reusable document profiles with checked batch extraction - #6
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Recurring labeled forms currently require agents to rediscover extraction rules or write scripts. Extend
extractwith a saved local JSON profile and one batch call: exact document anchors, required unique labels and text/decimal/date checks. Failed rows need review and withhold partial fields; review rows lead bounded previews. Optional complete JSONL reports preserve exact strings and protect documents, profiles and existing outputs. No training, inference dependency, telemetry or extra MCP tool.Prepare context 0.3.0 with examples, public install/discovery metadata and exact-artifact checks. Runtime 0.1.1 stays unchanged. Source tests cover real PDF/DOCX, CLI/MCP, drift, ambiguous/missing fields, limits, output safety and exact profile hashes. Local tests: 253 pass; clean tarball install checks pass. CI additionally verifies supported Node/OS combinations, containers and release artifacts.
Evidence deliberately includes overhead: the correct prepared script is faster; tiny documents are a poor fit. One saved-profile agent pair after fixing an extra call reported $0.1304 / 16.45s versus $0.2007 / 33.83s with built-ins, but setup is excluded and cache state differs. Both initial and corrected trials are retained. No universal savings claim. Shortened descriptions keep the serialized tool catalog slightly smaller than 0.2.0.