File Search

Intelligent file recognition and search is one of SingClaw's core capabilities for multi-file scenarios. It allows users to upload multiple files and then quickly locate target files, parse file structures, and identify field information through natural language commands — solving the pain points of "too many files, hard to locate, hard to understand" in multi-file management. It is the user's "intelligent file manager."

Core Functionality Flow & Experience

The example scenario clearly shows the full flow of this feature:

1. Batch File Ingestion — Users can upload multiple related files at once (e.g., Taobao order-related files). All files are managed in SingClaw's workspace without manual sorting or categorization.
2. Natural Language File Location — Users simply ask in natural language (e.g., "which file is related to orders?"), and the system automatically scans all uploaded files to quickly match and locate the target — no need to manually open and check each one.
3. Automatic File Structure Parsing — After locating the target file, the system automatically parses its core information, including file path, contained fields, and field meanings, presented in a structured format so users can quickly understand the content without opening the file.

---

Core Capability Details

1. Multi-File Intelligent Matching & Location

• Semantic File Search: Supports finding files by business semantics rather than file names — even with non-standard naming, the system can automatically match relevant files based on business keywords in the user's query (e.g., "orders," "users," "products").
• Multi-File Batch Filtering: Supports managing multiple files simultaneously and quickly filtering those relevant to the user's needs — especially useful for scenarios with multiple related data files (e.g., orders, users, products, traffic tables).
• File Source Traceability: Clearly marks the path and source of target files for easy subsequent access and use.

2. Automatic File Structure Parsing & Field Recognition

Using the order file parsing example, the system can automatically identify and annotate core field information:

Field Category	Example Fields	Parsing Description
Order Identifier	order_id	Identified as "Order Number" with clear field meaning
User Information	buyer_nickname / buyer_id	Identified as "Buyer Information" with related fields grouped
Product Information	product_name	Identified as "Product Name" with field purpose annotated
Transaction Data	price / quantity / actual_payment	Identified as "Price, Quantity, Actual Payment" with business meaning explained
Time Nodes	order_time / payment_time / shipping_time	Identified as "Order/Payment/Shipping Time" with time fields grouped
Status Information	logistics_status / order_status	Identified as "Logistics/Order Status" with status fields annotated

Results are presented as a structured list, allowing users to quickly understand the file's business purpose and data structure without opening it — significantly reducing file comprehension costs.

3. Context Memory & Multi-Turn Interaction

• Session-Based File Memory: Powered by the SingMemory module, the system remembers all files uploaded in the current session. Users can issue search commands directly without re-uploading or re-explaining file information.
• Multi-Turn File Interaction: Supports follow-up questions based on search results, e.g., "what order status fields are in this file?" or "show me another file related to users" — the system continuously responds to multi-turn file-related commands.

---

Typical Use Cases

1. Multi-File Data Processing: When working with multiple data files (orders, users, products, etc.) simultaneously, quickly locate target files to avoid manual searching through large file sets and improve analysis efficiency.
2. Quick Understanding of Unfamiliar Files: When receiving unfamiliar files (e.g., historical data files, third-party exports), quickly parse the file structure and field meanings to reduce comprehension costs.
3. Pre-Analysis for Cross-File Joins: Before performing multi-file join analysis, use this feature to confirm each file's fields and purpose, clarify analysis logic, and avoid errors caused by unclear file structures.
4. File Management & Archiving: Semantically manage large numbers of files — find target files quickly through natural language, replacing traditional folder categorization and manual search.

---

Core Value Summary

SingClaw's intelligent file recognition and search essentially provides users with a "semantic intelligent file manager." It solves the pain points of traditional file management — reliance on file names, low search efficiency, and difficulty understanding unfamiliar files — enabling users to quickly locate files and parse structures through natural language, laying the foundation for subsequent data analysis and business processing, and dramatically improving efficiency in multi-file scenarios.