EdgeQuake Feature Registry

This file maintains traceability between code features and business requirements.

Index

Feature ID	Description	Status	Spec / Issue
FEAT-0001	Tenant Workspace Quota Management	Completed	SPEC-0001 / #133
FEAT-0002	Knowledge Injection (Glossaries & Synonyms)	Completed	SPEC-0002 / #131
FEAT-0003	Explainability	Planned	SPEC-0003 / #128
FEAT-0004	Graph Edge Labels	Planned	SPEC-0004 / #91
FEAT-0005	Custom Entity Configuration	Completed	SPEC-0005 / #85
FEAT-006	Unified Streaming Response Protocol	Completed	SPEC-006 / #56
FEAT-007	Vector Storage SQL Pre-Filtering	Completed	SPEC-007
FEAT-008	Explicit Provider/Model Transparency in UI	Completed	MISSION-01 / v0.9.19
FEAT-009	Document Deletion Correctness	Completed	MISSION-02 / v0.9.19
FEAT-010	Configurable PDF Parser Backend (Vision/EdgeParse)	Completed	MISSION-03 / v0.10.0

Feature Definitions

FEAT-0002 — Knowledge Injection

Issue: #131
Spec: specifications/0002_knowledge_injection_issue_131
Released: v0.8.0 (2026-04-03)
Status: ✅ Completed

Problem: Domain-specific acronyms (OEE, NLP) and synonyms are unknown to the embedding model. Queries for “OEE” miss documents that say “Overall Equipment Effectiveness”, degrading retrieval quality.

Solution: Workspace owners inject glossary definitions as named entries. These are processed through the standard entity-extraction pipeline, enriching the knowledge graph. At query time, injection entities expand the query terms. Injection entries are never shown as source citations.

API Surface:

PUT /api/v1/workspaces/:id/injection — create/replace text injection
POST /api/v1/workspaces/:id/injection/upload — upload file injection
GET /api/v1/workspaces/:id/injection — list all entries
GET /api/v1/workspaces/:id/injection/:injection_id — get detail
PATCH /api/v1/workspaces/:id/injection/:injection_id — update name/content
DELETE /api/v1/workspaces/:id/injection/:injection_id — delete + cascade cleanup

UI: /knowledge page with list, add dialog (text/file tabs), detail page, inline edit, delete confirmation.

Test Coverage: 1 000+ line Rust E2E suite + 5 Playwright browser tests.

FEAT-0005 — Custom Entity Configuration

Issue: #85
Spec: specifications/0005_custom_entity_config_issue_85
Released: v0.8.0 (2026-04-03)
Status: ✅ Completed

Problem: Default generic entity types are insufficient for domain-specific corpora (manufacturing, healthcare, legal, research, finance), reducing extraction recall and graph quality.

Solution: Workspace creation supports entity_types with preset-driven and custom configuration. Types are normalized and stored in workspace metadata, then automatically injected into extraction prompts per workspace.

Capabilities:

Workspace-scoped entity_types in create-workspace API payload.
Validation and normalization: trim, uppercase, space/hyphen to underscore, dedupe, max 50.
Pipeline fallback to server defaults when no custom entity types are configured.
Frontend selector with presets (General, Manufacturing, Healthcare, Legal, Research, Finance) and custom add/remove chips.
Workspace detail page displays configured entity types.
Full i18n labels in en/fr/zh.

API Surface:

POST /api/v1/tenants/:tenant_id/workspaces accepts entity_types.
Workspace response surfaces entity_types (from metadata JSONB).

Test Coverage:

Rust integration coverage for workspace request/metadata threading.
Playwright coverage for selector UX edge cases and workspace detail display.

FEAT-010 — Configurable PDF Parser Backend

Spec: mission/03-pdf-parser.md
Released: v0.10.0 (2026-04-10)
Status: ✅ Completed

Problem: Vision-only PDF extraction is expensive, slower on digital-native PDFs, and unnecessarily dependent on an LLM for documents that already contain structured text.

Solution: EdgeQuake now supports two runtime PDF extraction backends:

vision for scanned, image-heavy, or layout-complex PDFs.
edgeparse for fast CPU-only extraction of digital-native PDFs.

Resolution order:

Per-upload multipart override pdf_parser_backend
Workspace default workspace.pdf_parser_backend
Environment variable EDGEQUAKE_PDF_PARSER_BACKEND
Fallback default vision

Capabilities:

New edgequake-pdf abstraction crate with a backend strategy pattern.
EdgeParse integration via edgeparse-core without temp files.
Workspace-level default parser setting on the workspace configuration page (/workspace and /w/[slug]/workspace).
Per-upload parser selection in the document upload flow via the Parser for this upload selector, including a Workspace Default option.
Extraction lineage includes parser method and low-content warnings for image-only PDFs.
EdgeParse markdown is sanitized before persistence to remove embedded NUL bytes that PostgreSQL rejects as invalid UTF-8 payload.
Storage now records extraction_method = edgeparse.

Operational note:

EdgeParse does not auto-fallback to Vision. If output is low-content, the UI surfaces a warning so the user can explicitly retry with Vision.

Last Updated: 2026-04-10 Total Features: 10