Parent epic: #8 (post-v0.1 backlog)
Context
§1.4 records that screenshots were promised while OCR was excluded, a real mismatch, and resolved it by removing screenshots and scanned documents from the v0.1 promise and listing them as a future parser class. §5.2 excludes OCR, screenshots, images, audio, video, and PPTX.
Post-v0.1. The v0.1 behavior (refuse clearly) must remain until this ships.
Spec references
§1.4 feedback audit, §5.2 Explicitly out of scope, §8.7 (optional parser adapters), §24.4 parsing quality risk.
Scope
In: an evaluation of the visual/scanned artifact parser class. OCR for scanned PDFs and images, and PPTX, including dependency implications and quality expectations.
Out: shipping a parser in this issue; any change to v0.1's clear refusal behavior.
Design notes
- The blocking constraint is §6.1: no native dependency, no Python sidecar, no compiler toolchain in the core install. Any OCR engine must therefore be an optional package with the same explicit-install contract as the semantic package (§5.6), or be rejected.
- PPTX is a separate, easier question (an OOXML format like DOCX/XLSX) and may be viable much sooner than OCR, the spike should not bundle their fates together.
- Quality expectations must be honest: OCR output is lossy, and §4.4/§24.4 require visible warnings rather than silent degradation.
Acceptance criteria
Tests / verification
- Prototype determinism check on a small scanned corpus if an engine is shortlisted.
- Findings recorded as an ADR entry.
Dependencies
Blocked by the v0.1 release (#102).
Definition of Done
The visual-artifact question is answered with evidence, and PPTX is prioritized independently of OCR.
Parent epic: #8 (post-v0.1 backlog)
Context
§1.4 records that screenshots were promised while OCR was excluded, a real mismatch, and resolved it by removing screenshots and scanned documents from the v0.1 promise and listing them as a future parser class. §5.2 excludes OCR, screenshots, images, audio, video, and PPTX.
Post-v0.1. The v0.1 behavior (refuse clearly) must remain until this ships.
Spec references
§1.4 feedback audit, §5.2 Explicitly out of scope, §8.7 (optional parser adapters), §24.4 parsing quality risk.
Scope
In: an evaluation of the visual/scanned artifact parser class. OCR for scanned PDFs and images, and PPTX, including dependency implications and quality expectations.
Out: shipping a parser in this issue; any change to v0.1's clear refusal behavior.
Design notes
Acceptance criteria
Tests / verification
Dependencies
Blocked by the v0.1 release (#102).
Definition of Done
The visual-artifact question is answered with evidence, and PPTX is prioritized independently of OCR.