SLM JSON Cleaner
Lightweight local JSON text sanitizer and structural repair tool that formats malformed, truncated, or unquoted LLM outputs offline on standard CPU.
02 / Standalone OS Harness Installers, CLI & Python SDK
In SLMAgents 2.0, SLM JSON Cleaner executes inside the 10-Plane Pure-CPU Agent Harness (slm_harness). Download a self-extracting Individual Agent + Harness OS Installer for macOS, Windows, or Linux (bundling slm_json_cleaner + slm_harness + the ./slm CLI), install the full 26-agent desktop suite, or install via Python pip:
1. Individual SLM JSON Cleaner + 10-Plane Harness Installers (All OSes)
slm_json_cleaner + slm_harness + ./slm CLIslm_json_cleaner + slm_harnessslm_json_cleaner + slm_harness + ./slm CLI2. Full 26-Agent Studio & Desktop OS Installers
3. 10-Plane Harness CLI (./slm) & Python pip Installation
# Option A: Install via PyPI or editable source with the 10-Plane Harness $ pip install slm-json-cleaner $ pip install -e ./slm_harness -e ./slm_json_cleaner # Option B: Run directly through the Unified 10-Plane Pure-CPU Harness CLI (./slm) $ cat broken.json | ./slm ask "Repair and normalize JSON" --schema '{"type":"object","properties":{"status":{"type":"string"}},"required":["status"]}' $ ./slm chat # Interactive Rich TUI with 18 slash commands (/trace, /models, /arena, /undo) $ ./slm serve # 100% Offline Web AI Studio + OpenAI /v1 API at http://127.0.0.1:7860
03 / 10-Plane Pure-CPU Harness Integration (slm_harness)
Rather than executing as an isolated prompt wrapper, SLM JSON Cleaner is natively orchestrated by the SLMAgents 2.0 10-Plane Pure-CPU Harness (via slm_harness.runtime.legacy_bridge and CascadeRouter). Every call benefits from hybrid llama.cpp GGUF (Q4_K_M) + Quantized INT4/INT8 CPU execution, watermarked LRU memory governance, grammar-constrained decoding, and deterministic verification at $0 API cost:
| Harness Plane | Subsystem | How It Powers SLM JSON Cleaner on Pure CPU |
|---|---|---|
| Plane 6 | CascadeRouter (<35ms) |
Stage 1 (Malformed JSON / Schema Payload Detection) & Stage 2 (Brace/Quote Signature) |
| Plane 2 & 5 | MemoryArena & Hybrid CPU Runtime |
Specialist Role: primary (Qwen3-4B GGUF Q4_K_M / Qwen3.5-0.8B INT4). Managed inside Watermark LRU MemoryArena (lite 2.6 GB / balanced 4.7 GB / power 8.2 GB) with zero-copy session sharing and automatic OOM tier step-down (8B -> 4B -> 3B -> 0.8B). |
| Plane 7 & 8 | Deterministic Verifier & Governance | Plane 2 LlamaGrammar (GBNF) Constrained Decoding + Deterministic RFC 8259 JSON Repair & Schema Validator |
| Plane 1, 3 & 4 | Kernel, Artifact Scratchpad & 4-Tier Memory | SQLite WAL CheckpointStore (/fork, /rewind, /undo), content-addressed artifact://<sha12> handles to prevent prompt bloat, and hybrid SQLite FTS5 BM25 + HNSW vector memory. |
04 / Source Checkout (Git Sparse-Checkout)
Clone this agent alongside the 10-Plane Pure-CPU Harness (slm_harness) from the monorepo using Git sparse-checkout — no need to download all 26 agents:
Option 1 — Sparse Checkout with 10-Plane Harness (Recommended)
# 1. Create and enter a new directory $ mkdir slm_json_cleaner_workspace && cd slm_json_cleaner_workspace # 2. Initialise empty git repo and add remote $ git init $ git remote add origin https://github.com/t00114218-stack/SLMAgents.git # 3. Enable sparse-checkout for both slm_harness and slm_json_cleaner $ git sparse-checkout init --cone $ git sparse-checkout set slm_harness slm_json_cleaner # 4. Pull the 10-Plane Harness and agent source $ git pull origin main
Option 2 — Full Repository Clone
$ git clone https://github.com/t00114218-stack/SLMAgents.git $ cd SLMAgents $ ./slm doctor
Note: After checkout, install the harness and agent in editable mode with pip install -e ./slm_harness -e ./slm_json_cleaner or download the pre-packaged OS Harness Installer above.
05 / Configuration & API Reference
Constructor Parameters
| Parameter | Type / Default | Description |
|---|---|---|
| model_path | str | None | Explicit path to GGUF (Q4_K_M) or quantized ONNX model directory. Defaults to zero-copy sharing via the 10-Plane Harness MemoryArena. |
| cache_dir | str | None | HF model directory path. Also settable via SLM_JSON_CLEANER_CACHE_DIR. |
| n_ctx | int | 2048 | Context window size in tokens. Also settable via SLM_JSON_CLEANER_N_CTX. |
| n_threads | int | 4 | CPU threads for the 10-Plane Pure-CPU Harness (auto-tuned to physical P-cores). Also settable via SLM_JSON_CLEANER_N_THREADS. |
| system_prompt | str | None | Optional custom system prompt instructions overriding the default template. |
| user_input | str | None | Optional additional user-supplied target parameters or variables. |
clean_json() Parameters
| Parameter | Type / Default | Description |
|---|---|---|
| malformed_text | str | Required. Unstructured or broken JSON input string to repair. |
| schema_dict | dict | Required. A dictionary mapping keys and types required in the final output (e.g. {"name": "string", "age": "number"}). |
| system_prompt | str | None | Optional custom system prompt instructions overriding the default template. |
| user_input | str | None | Optional additional user-supplied target parameters or variables. |
from slm_json_cleaner.json_cleaner import SLMJSONCleaner
cleaner = SLMJSONCleaner()
# Truncated, nested, and conversational malformed input log:
broken_input = """
Raw logs:
{
"project_name": "Antigravity Pipeline",
"build_status": "success",
"metrics": {
"duration_seconds": 124.5,
"test_count": 48,
"failed_tests": 0,
"coverage": "98.5%
},
"contributors": [
{"name": "Alice", "role": "lead"},
{"name": "Bob", "role": "reviewer"
],
"releases": [
"v1.0", "v1.1",
[LOG EXPIRED MID-TOKEN]
"""
schema = {
"project_name": "string",
"build_status": "string",
"metrics": {
"duration_seconds": "number",
"test_count": "number",
"failed_tests": "number",
"coverage": "string"
},
"contributors": [
{"name": "string", "role": "string"}
],
"releases": ["string"]
}
parsed, success = cleaner.clean_json(broken_input, schema)
print(f"Success: {success}")
print(parsed)
JSON Recovery & Repair Capabilities
The sanitization model operates as a high-speed pattern matching engine. It cleans common structural errors caused by early tokens cutoff or memory limits:
- Auto-closure: Repairs unclosed brackets
}or list braces]. - Quote repair: Fixes missing double-quotes on JSON keys and string values.
- Regex brace fallback: Includes a backup regex matcher to extract matching curly braces if standard JSON parsing triggers an exception.
Target Schema Enforcement
By supplying a target schema dictionary, you guarantee that the returned attributes match the expected data types. The repaired response returns exactly structured JSON data:
{
"project_name": "Antigravity Pipeline",
"build_status": "success",
"metrics": {
"duration_seconds": 124.5,
"test_count": 48,
"failed_tests": 0,
"coverage": "98.5%"
},
"contributors": [
{ "name": "Alice", "role": "lead" },
{ "name": "Bob", "role": "reviewer" }
],
"releases": [ "v1.0", "v1.1" ]
}