arglib 0.1.9

Build a general-purpose, batteries-included Python library for creating, importing, analyzing, and reasoning over argument graphs derived from text and multimodal evidence.

ArgLib

ArgLib is a batteries-included Python library for creating, importing, analyzing, and reasoning over argument graphs derived from text and multimodal evidence.

Highlights

• Canonical ArgumentGraph model with provenance-aware nodes and relations.
• Warrant-gated scoring with claim, warrant, and gate scores.
• Diagnostics for cycles, components, reachability, and degree stats.
• JSON IO with schema validation and Graphviz DOT export.
• CLI tools for DOT, diagnostics, and validation.
• Argument bundles for higher-level reasoning and credibility propagation scoring.
• Evidence cards and supporting documents for evidence pipelines.
• Deterministic evidence scoring and edge validation helpers (LLM adapters planned).

Install

python -m pip install arglib

Quickstart

from arglib.core import ArgumentGraph
from arglib.reasoning import compute_credibility

graph = ArgumentGraph.new(title="Parks")
c1 = graph.add_claim("Green spaces reduce urban heat.", type="fact")
c2 = graph.add_claim("Cities should fund parks.", type="policy")
graph.add_support(c1, c2, rationale="Cooling improves health", gate_mode="OR")

credibility = compute_credibility(graph)
scores = credibility.final_scores

Evidence and scoring

from arglib.ai import score_evidence, validate_edges

scores = score_evidence(graph)
edge_report = validate_edges(graph)

Bundles and credibility propagation

from arglib.reasoning import compute_credibility

bundle = graph.define_argument([c1, c2], bundle_id="arg-1")
cred = compute_credibility(graph)

CLI examples

arglib dot path/to/graph.json
arglib diagnostics path/to/graph.json --validate
arglib validate path/to/graph.json

Development

This repo uses uv for dependency management.

uv sync
scripts/check.sh

Documentation

Full docs and guides are available at https://vasanthsarathy.github.io/arglib/.