Document Review Agent

A powerful Dify plugin providing comprehensive AI-powered document review capabilities for various types of documents including tender documents, official documents, contracts, and materials, with support for non-compliant document detection. Supports intelligent document parsing, rule-based auditing, risk aggregation, and annotated document generation with professional-grade quality and flexible configuration options.

Version Information

• Current Version: v0.0.4
• Release Date: 2026-05-16
• Compatibility: Dify Plugin Framework
• Python Version: 3.12

Version History

• v0.0.4 (2026-05-16):
  • Added multi-language progress prompts: supports Chinese and English progress and error messages during audit execution
  • Bug Fix
• v0.0.3 (2026-05-03):
  • Optimized document revision feature: improved text matching logic across multiple text blocks for more accurate revision results
• v0.0.2 (2026-04-12):
  • Added integrated slice audit tool doc-slice-audit (parse -> load rules -> audit -> aggregate -> annotate -> revise)
  • Added integrated simple/full-text audit tool doc-audit for short document single-loop auditing
  • Added template slice audit tool doc-slice-audit-template with required template_file and optional rules_file
  • Added template full-text audit tool doc-audit-template with required template_file and optional rules_file
  • Added template comparators: template_chunk_auditor.py and template_doc_auditor.py
  • Added template risk code normalization to template-0001 style and aligned output fields for aggregation/annotation
  • Improved no-risk handling in doc_annotator (returns original reviewed file with annotation_count=0 instead of failing)
  • Reorganized provider tool exposure and YAML definitions around integrated top-level tools
• v0.0.1 (2026-04-05): Initial release with local document review capabilities

Quick Start

1. Install plugin in your Dify environment

2. Download Rules Template and Sample Files: https://github.com/sawyer-shi/awsome-dify-agents/blob/master/src/doc-review-agent/agent_test_files/review_rules_research_en.csv

3. Configure your LLM model settings. Also note: To prevent timeout, you can modify the parameter PLUGIN_MAX_EXECUTION_TIMEOUT to increase processing time!!!

4. Upload your document and start the review process. Results are as follows:

Key Features

• Four Integrated Audit Tools: Slice/non-template, full-text/non-template, slice/template, and full-text/template workflows

• Template Baseline Review: Template-based findings use normalized risk codes like template-0001 for consistent downstream tagging

• Hybrid Rule + Template Aggregation: Optional rules_file can run together with template audit and merge into one unified risk payload

• Structured Risk Pipeline: Audit -> aggregation -> annotation -> revision with consistent data schema across workflows

• High-Quality Output Files: Reviewed (annotated) and revised .docx outputs with configurable JSON/file output modes

• Flexible Control Knobs: Slice strategy, audit strategy, merge policy, merge strategy, language, and output settings

• No-Risk Safe Handling: When no risks are found, the workflow returns a valid reviewed file instead of failing

• Multi-Language Reasoning: Supports zh/en/ja/ko/es/fr/de/pt/ru/ar outputs

  

Core Features

1) Doc Slice Audit (doc-slice-audit)

Integrated non-template slice auditing for larger documents.

• Required: model_config, upload_file, rules_file
• What it does (6 steps):
  1. Document slicing
  2. Rule loading
  3. Chunk auditing
  4. Risk aggregation
  5. Document annotation
  6. File revision
• Best for: contracts/tenders where chunk-level analysis is preferred

2) Doc Audit (doc-audit)

Integrated non-template full-text auditing for short documents.

• Required: model_config, upload_file, rules_file
• What it does (6 steps):
  1. Load review document
  2. Rule loading
  3. Full-text rule audit
  4. Risk aggregation
  5. Document annotation
  6. File revision
• Best for: shorter documents where whole-text context is important

3) Doc Slice Audit Template (doc-slice-audit-template)

Integrated template-based slice auditing.

• Required: model_config, upload_file, template_file
• Optional: rules_file (runs rule audit + template audit together when provided)
• What it does (8-step pipeline):
  1. Slice review document
  2. Slice template document
  3. Rule loading (optional input; step kept in progress output)
  4. Rule-based chunk audit (runs when rules_file is provided, otherwise marked as skipped)
  5. Template chunk comparison audit
  6. Risk aggregation
  7. Document annotation
  8. File revision
• Output semantics: template findings use normalized codes (template-0001, template-0002, ...), severity from LLM (high|medium|low)

4) Doc Audit Template (doc-audit-template)

Integrated template-based full-text auditing.

• Required: model_config, upload_file, template_file
• Optional: rules_file (runs rule audit + template audit together when provided)
• What it does (8-step pipeline):
  1. Load review document
  2. Load template document
  3. Rule loading (optional input; step kept in progress output)
  4. Rule-based full-text audit (runs when rules_file is provided, otherwise marked as skipped)
  5. Full-text template comparison audit
  6. Risk aggregation
  7. Document annotation
  8. File revision
• Best for: short-form baseline checks against a model template

Shared Output and Controls

• JSON output: summary_only or detailed
• File output: revised only, or reviewed + revised
• Revision behavior: choose merge strategy and whether to apply revisions back to source text
• No-risk behavior: returns a valid reviewed file with annotation_count=0

Technical Advantages

• LLM-Powered Analysis: Leverages advanced LLM models for intelligent document understanding
• Rule-Based Auditing: Flexible rule system for customizable review criteria
• Chunk-Based Processing: Efficient handling of large documents through intelligent slicing
• Risk Deduplication: Smart aggregation to eliminate redundant findings
• Annotated Output: Professional document output with clear risk indicators
• Multi-Format Support: Optimized for docx format with extensibility for other formats
• Configurable Audit Levels: Support for strict and lenient auditing modes
• Real-Time Processing: Efficient workflow for timely document review

Requirements

• Python 3.12
• Dify Platform access
• Configured LLM model
• Required Python packages (installed via requirements.txt):
  • dify_plugin>=0.5.0
  • python-docx>=1.1.2
  • openpyxl>=3.1.5

Installation & Configuration

1. Install required dependencies:

   pip install -r requirements.txt
   

2. Configure your LLM model in plugin settings

3. Install plugin in your Dify environment

Usage

Choose the Right Tool

A) Non-template Slice Audit

Use doc-slice-audit when you have a rule file and need chunk-level review.

• Required: model_config, upload_file, rules_file
• Recommended options: slice_strategy, max_chunk_chars, merge_policy, output_language

B) Non-template Full-Text Audit

Use doc-audit when you have a rule file and the document is short enough for full-text auditing.

• Required: model_config, upload_file, rules_file
• Recommended options: audit_strategy, merge_policy, output_language

C) Template Slice Audit

Use doc-slice-audit-template when template compliance is required at chunk level.

• Required: model_config, upload_file, template_file
• Optional: rules_file for hybrid rule + template audit
• Notes: template findings are normalized to template-0001 style risk codes

D) Template Full-Text Audit

Use doc-audit-template when template compliance is required for the full document.

• Required: model_config, upload_file, template_file
• Optional: rules_file for hybrid rule + template audit

Typical Output

• A JSON summary (or detailed JSON if enabled)
• A reviewed .docx (annotations)
• A revised .docx (merged or applied revisions)

Supported Document Formats

• Input: .docx (Microsoft Word)
• Output: .docx (Microsoft Word with annotations)

Notes

• Document parsing is optimized for docx format
• Chunk size can be adjusted based on document complexity
• Audit level affects the strictness of rule application
• Risk aggregation uses intelligent deduplication to avoid redundant findings
• Annotation style currently supports comment-based annotations
• Large documents are processed efficiently through chunking
• All tools require a configured LLM model for operation

Developer Information

• Author: https://github.com/sawyer-shi
• Email: sawyer36@foxmail.com
• License: Apache License 2.0
• Source Code: https://github.com/sawyer-shi/dify-plugins-doc-review-agent
• Support: Through Dify platform and GitHub Issues

License Notice

This project is licensed under Apache License 2.0. See LICENSE file for full license text.

---

Ready to review your documents with AI-powered intelligence?