aiex-cli 0.0.5-beta.6 → 0.0.6-beta.2

package/README.md

@@ -202,23 +202,13 @@ aiex completion fish | source
 
 <br>
 
-## 📄 Large Document Processing
-
-`aiex` uses a unified text extraction pipeline for both short and very large documents. Source files are converted to text or Markdown first; images are converted with OCR before structured extraction.
-
-- **Token-Aware AST Splitting**: Parses structural Markdown elements (headings, paragraphs, lists) using an AST-based parser (`marked.lexer`) and splits them using precise token counters (`js-tiktoken`). Active heading hierarchies are tracked and prepended to each chunk as context. Short documents run through the same pipeline as a single chunk.
-- **Concurrency Limiting**: To respect strict model rate limits, chunk extractions are processed in parallel with a strict concurrency limit (capped at 2 concurrent requests).
-- **Candidate & Evidence Merging**: Chunk results are merged into schema-shaped candidates, with evidence coverage used to select scalar conflicts and preserve traceability.
-- **Schema Validation & Correction**: Merged output is validated against the JSON Schema. When correction is needed, the corrected output is rechecked against evidence before being written.
-
-<br>
-
 ## 🔧 AI Configuration
 
 aiex works with any OpenAI-compatible API provider. Configure in the Web UI (AI Settings panel):
 
 - **Provider** — Set your base URL and API key
 - **Models** — Add models with vision and/or structured output capabilities
+- **Documents** — Choose a PDF converter (`unpdf`, `mineru`, `mineru_api`, or `external`); image input automatically uses a vision model when available, otherwise system OCR on supported platforms
 - **Prompts** — Customize system and user prompt templates with `{schema}` and `{text}` placeholders
 - **Integrations** — Optionally connect Notion from AI Settings; use Connect & Map to bind a schema to an existing Notion data source