Help System (how_to_use)

On-demand documentation for reduced token consumption

Remembrances includes a built-in help system called how_to_use that provides on-demand documentation for all tools. This design reduces initial token consumption by approximately 85%, making your AI interactions more efficient and cost-effective.

Why how_to_use?

Traditionally, AI agents receive full documentation for all available tools at the start of each conversation. For Remembrances with its 37+ tools, this means loading ~15,000+ tokens of documentation before any actual work begins.

The how_to_use approach changes this:

  • Minimal initial context: Each tool has a 1-2 line description
  • On-demand details: Full documentation loaded only when needed
  • Better focus: AI agents see relevant documentation when they need it

Token Savings

MetricTraditionalWith how_to_useSavings
Initial context~15,000 tokens~2,500 tokens~83%
Per conversationSame every timeOnly what’s neededVaries

How to Use

Get Complete Overview

Ask your AI to call how_to_use() without parameters:

how_to_use()

This returns a high-level overview of all tool categories:

  • Memory tools (facts, vectors, graph)
  • Knowledge base tools
  • Code indexing tools
  • Events tools

Get Group Documentation

For detailed information about a category of tools:

how_to_use("memory")

Available groups:

  • memory – Facts, vectors, and graph operations
  • kb – Knowledge base document tools
  • code – Code indexing and search tools
  • events – Event logging and search tools

Get Specific Tool Documentation

For complete documentation on a single tool:

how_to_use("remembrance_save_fact")
how_to_use("code_semantic_search")
how_to_use("save_event")

This returns:

  • Full description
  • All parameters with types and descriptions
  • Usage examples
  • Related tools

For AI Agents

When your AI agent encounters an unfamiliar tool or needs more information, it can use how_to_use to get exactly the documentation it needs. This pattern:

  1. Reduces context overhead: Only loads documentation when needed
  2. Improves accuracy: Fresh, focused documentation at the point of use
  3. Saves costs: Fewer tokens means lower API costs

Example Workflow

Instead of having all tool documentation loaded upfront, your AI agent can:

  1. See the brief tool descriptions in its initial context
  2. When it needs to use a specific tool, call how_to_use("tool_name")
  3. Get detailed parameters and examples
  4. Proceed with the actual tool call

Best Practices

For Users

  • Let your AI discover tools naturally through how_to_use()
  • If your AI seems confused about a tool, suggest it calls how_to_use("tool_name")
  • Start sessions by having the AI check how_to_use() for an overview if it’s unfamiliar with Remembrances

For AI Agents

  • Use how_to_use() at the start of complex tasks to understand available capabilities
  • Look up specific tools before using unfamiliar functionality
  • Reference group documentation when working with a category of operations