SLM (Small Language Model) — MCP Tool Specification
Version: v1.1.9 | Status: Active | Last Updated: March 2026
Overview
This document defines the MCP tools exposed by the slm module.
These tools are auto-discovered by the PAI MCP bridge via the @mcp_tool decorator
in mcp_tools.py and surfaced as part of the ~303 dynamic tools available to Claude.
The SLM module provides a tiny transformer-based language model for experimentation, supporting both token generation and forward-pass logit analysis.
Auto-Discovery
| Property | Value |
|---|---|
| Discovery method | @mcp_tool decorator scan |
| Namespace | slm |
| Trust default | Safe |
| PAI bridge | src/codomyrmex/agents/pai/mcp/ |
Tool Reference
slm_generate
Description: Generate tokens from a tiny language model. Trust Level: Safe Category: generation
Parameters:
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
prompt_tokens | list[int] | No | [1, 2, 3] | List of integer token IDs for the prompt |
max_new_tokens | int | No | 10 | Number of new tokens to generate |
vocab_size | int | No | 100 | Vocabulary size |
d_model | int | No | 32 | Model dimension |
n_heads | int | No | 2 | Number of attention heads |
n_layers | int | No | 1 | Number of transformer layers |
seed | int | No | 42 | Random seed for reproducibility |
Returns: dict — Dictionary with prompt (original tokens), generated (new tokens), full_sequence (concatenation), vocab_size, model_params (dict), and status.
Example:
from codomyrmex.slm.mcp_tools import slm_generate
result = slm_generate(prompt_tokens=[1, 5, 10], max_new_tokens=20, vocab_size=200)
Notes: Creates a fresh model instance on each call with random weights. Useful for testing generation pipelines, not for meaningful text output.
slm_forward
Description: Run a forward pass through the SLM and return logit statistics. Trust Level: Safe Category: analysis
Parameters:
| Parameter | Type | Required | Default | Description |
|---|---|---|---|---|
batch_size | int | No | 1 | Number of sequences in the batch |
seq_len | int | No | 8 | Length of each sequence |
vocab_size | int | No | 100 | Vocabulary size |
d_model | int | No | 32 | Model dimension |
n_heads | int | No | 2 | Number of attention heads |
n_layers | int | No | 1 | Number of transformer layers |
seed | int | No | 42 | Random seed for reproducibility |
Returns: dict — Dictionary with output_shape, logit_mean, logit_std, logit_min, logit_max, and status.
Example:
from codomyrmex.slm.mcp_tools import slm_forward
result = slm_forward(batch_size=2, seq_len=16, d_model=64, n_heads=4)
Notes: Generates random token IDs internally. Returns statistical summaries of the logit tensor rather than the full output.
Integration Notes
- Auto-discovered: Yes (via
@mcp_toolinmcp_tools.py) - Trust Gateway: All tools are safe — no trust check required
- PAI Phases: BUILD (model prototyping), VERIFY (output shape/statistics validation)
- Dependencies:
numpy,slm.model.SLM,slm.model.SLMConfig