qwen3-repo 0.1.0

Architecture-aware repo-to-context scaffold for Qwen3 and Qwen3.6

qwen3-repo

Architecture-aware repo-to-context scaffold for Qwen3 and Qwen3.6. Ingests a GitHub repository into a dependency-ordered context pack, runs an agentic coding loop via any OpenAI-compatible endpoint, and includes NL2Repo / SWE-bench evaluation runners.

Why ordering matters

Qwen3.6's hybrid architecture (Gated DeltaNet + Gated Attention) processes three out of four layers with linear attention and a fixed-size recurrent state. Placing definitions before their dependents should help the recurrent state accumulate core types and interfaces before call sites reference them.

Standard Qwen3 models use full attention and can look back to any position, so ordering matters less -- but dependency-aware packing still avoids wasting context on low-value files and keeps related code together.

Qwen's NL2Repo evaluations (score: 36.2) were run via Claude Code (temp=1.0, top_p=0.95, max_turns=900). This scaffold provides an open-source alternative with architecture-aware context formatting.

Installation

pip install qwen3-repo

Quick start

Ingest a repository

from qwen3_repo import ingest_repo

# Works with any supported Qwen3 or Qwen3.6 model
context_pack, files, budget = ingest_repo(
    "https://github.com/user/repo",
    model="Qwen3-32B",
)
print(f"{len(files)} files, ~{budget.pack_budget:,} token budget")

Rank files by importance

from qwen3_repo import rank_files
from qwen3_repo.ingester import discover_files
from pathlib import Path

files = discover_files(Path("/path/to/repo"))
ranked = rank_files(files)
for r in ranked[:10]:
    print(f"{r.score:.2f}  {r.path}")

Detect vision encoder needs

from qwen3_repo import detect_vision_needs
from pathlib import Path

# Vision encoder is only relevant for Qwen3.6 (Qwen3 has no vision)
result = detect_vision_needs(Path("/path/to/repo"))
print(result["recommendation"])
# "Consider --language-model-only. Frees ~6 GB KV cache (~100K-150K additional tokens)."

Run the agentic scaffold

# Start a vLLM server first:
# vllm serve Qwen/Qwen3-32B --port 8000

python -m qwen3_repo.scaffold \
    --repo-path /path/to/repo \
    --task "Fix the failing test in test_auth.py" \
    --max-turns 900 \
    --temperature 1.0 \
    --top-p 0.95

Run NL2Repo evaluation

python -m qwen3_repo.eval.nl2repo \
    --tasks nl2repo_tasks.json \
    --api-url http://localhost:8000/v1 \
    --model Qwen/Qwen3.6-27B \
    --output-dir nl2repo_results

Compare Claude Code vs qwen3-repo

python -m qwen3_repo.bench_compare \
    --tasks comparison_tasks.json \
    --repo-path /path/to/repo \
    --markdown

Context ordering strategy

1. Role-based grouping: CONFIG -> TYPE_DEF -> CORE_LIB -> UTILITY -> FEATURE -> TEST -> DOC -> BUILD
2. Dependency-aware ordering: Topological sort within each group (Kahn's algorithm, importance as tiebreaker)
3. Budget trimming: Lowest-importance files dropped first; tests and docs trimmed before core code

Importance scoring signals

Signal	Weight	Description
Centrality	3.0	How many files import this file (log-scaled)
Role weight	2.0	TYPE_DEF > CONFIG > CORE_LIB > FEATURE > TEST > DOC
Recency	1.5	Exponentially decayed git last-modified (30-day half-life)
Coverage	1.0	Whether corresponding test files exist
Structural	0.8	Directory depth (root > deep nesting)
Size efficiency	0.5	Information density (tokens/byte)

Vision encoder decision (Qwen3.6 only)

Qwen3 models have no vision encoder. For Qwen3.6:

Repo contents	Vision encoder	Reason
No images	Disabled (--language-model-only)	Frees ~6 GB KV cache
Design mockups/screenshots	Enabled	Model needs to see design intent
SVG diagrams	Enabled	Vision helps with SVG understanding
Canvas/WebGL code	Enabled	Vision helps understand visual output
Only icons/favicons	Disabled	Low relevance, not worth KV cost

Supported models

Qwen3 (standard transformer, full attention)

Model	Context
Qwen3-235B-A22B	32K native, 128K extended
Qwen3-32B	32K native, 128K extended
Qwen3-30B-A3B	32K native, 128K extended
Qwen3-14B	32K native, 128K extended
Qwen3-8B	32K native, 128K extended
Qwen3-4B	32K native
Qwen3-1.7B	32K native
Qwen3-0.6B	32K native

Qwen3.6 (hybrid Gated DeltaNet + Gated Attention)

Model	Layers	Layout	Context
Qwen3.6-27B	64	16 x (3 x GDN + 1 x GA)	262K native, 1M extended
Qwen3.6-35B-A3B	40	10 x (3 x GDN + 1 x GA), MoE	262K native

License

Apache 2.0