polyglot-stream-heal (Python Native PyO3 Binding — PyPI)

Sub-millisecond incremental $O(n)$ Hierarchical Multi-Grammar Pushdown Transducer (HM-PDT) for streaming polyglot LLM output healing and cardinality-bounded layout partitioning.

• Author: David Charles Michael (davidchar@gmail.com)
• License: MIT
• Native Rust Performance in Python: Compiled via PyO3 + maturin (abi3-py39+), delivering $34.7\times$ lower median chunk latency ($6.4,\mu\text{s}$ vs. $222.2,\mu\text{s}$) and $267.6\times$ lower cumulative stream overhead at $32\text{ KB}$ than pure-Python json-repair, while additionally healing 14 Mermaid diagram dialects, HTML5 sandboxes, and wide Markdown tables.

1. Installation

pip install polyglot-stream-heal

Or install the pre-built wheel directly from dist/pypi/:

pip install dist/pypi/polyglot_stream_heal-0.1.0-cp39-abi3-manylinux_2_34_x86_64.whl

2. How to Implement & Verified Output

Runnable Code Snippet (python3 examples/demo_python.py)

import json
import polyglot_stream_heal as psh

print("=== 1. Stateful Streaming Transducer (Python PyO3) ===")
transducer = psh.IncrementalTransducer(max_table_cols=5, max_chart_bars=10)
stream_chunks = [
    "### Telemetry Stream\n```json\n",
    '{"service": "router", "active": tru, "p99_ms": 12.',
]
for chunk in stream_chunks:
    transducer.push_chunk(chunk)

active_seg = json.loads(transducer.snapshot_active_json())
print("Active Segment Modality:", active_seg["modality"])
print("Healed JSON Content    :", active_seg["content"])

print("\n=== 2. Context-Free Mermaid Dyck Node Healing ===")
raw_mermaid = "flowchart TD\n  subgraph Shard_A (Primary: US-East)\n    N1[Compute O(f(g(x)))] -- rpc --> N2[(Ledger (v2))]"
print(psh.heal_mermaid(raw_mermaid))

Exact Verified Output (examples/VERIFIED_OUTPUT.txt)

=== 1. Stateful Streaming Transducer (Python PyO3) ===
Active Segment Modality: json
Healed JSON Content    : {"active":true,"p99_ms":12.0,"service":"router"}

=== 2. Context-Free Mermaid Dyck Node Healing ===
flowchart TD
  subgraph sg_Shard_A__Primary__US ["Shard_A (Primary: US-East)"]
  N1["Compute O(f(g(x)))"] -->|rpc| N2[("Ledger (v2)")]
  end

3. Publishing to PyPI

Pre-Built Artifact Location

• dist/pypi/polyglot_stream_heal-0.1.0-cp39-abi3-manylinux_2_34_x86_64.whl

Option A: Automated 1-Command Publish via WSL2 Docker

1. Create a PyPI API token at https://pypi.org/manage/account/token/ (begins with pypi-...).
2. Run from the repository root:

   PYPI_API_TOKEN=pypi-your_token_here ./publish_packages.sh --pypi
   

Option B: Direct CLI Publish (twine or maturin)

python3 -m twine upload dist/pypi/polyglot_stream_heal-0.1.0-*.whl -u __token__ -p <YOUR_PYPI_API_TOKEN>