Thales Maciel 4d0081d1d0 Split aman.py into focused CLI and runtime modules

Break the old god module into flat siblings for CLI parsing, run lifecycle, daemon state, shared processing helpers, benchmark tooling, and maintainer-only model sync so changes stop sharing one giant import graph.

Keep aman as a thin shim over aman_cli, move sync-default-model behind the hidden aman-maint entrypoint plus Make wrappers, and update packaging metadata plus maintainer docs to reflect the new surface.

Retarget the tests to the new seams with dedicated runtime, run, benchmark, maintainer, and entrypoint suites, and verify with python3 -m unittest discover -s tests -p "test_*.py", python3 -m py_compile src/*.py tests/*.py, PYTHONPATH=src python3 -m aman --help, PYTHONPATH=src python3 -m aman version, and PYTHONPATH=src python3 -m aman_maint --help.

2026-03-14 14:54:57 -03:00

3.4 KiB

Raw Blame History

Developer And Maintainer Workflows

This document keeps build, packaging, development, and benchmarking material out of the first-run README path.

Build and packaging

make build
make package
make package-portable
make package-deb
make package-arch
make runtime-check
make release-check
make release-prep

make package-portable builds dist/aman-x11-linux-<version>.tar.gz plus its .sha256 file.
make release-prep runs make release-check, builds the packaged artifacts, and writes dist/SHA256SUMS for the release page upload set.
make package-deb installs Python dependencies while creating the package.
For offline Debian packaging, set AMAN_WHEELHOUSE_DIR to a directory containing the required wheels.

For 1.0.0, the manual publication target is the forge release page at https://git.thaloco.com/thaloco/aman/releases, using docs/releases/1.0.0.md as the release-notes source.

Developer setup

uv workflow:

uv sync
uv run aman run --config ~/.config/aman/config.json

pip workflow:

make install-local
aman run --config ~/.config/aman/config.json

Support and control commands

make run
make run config.example.json
make doctor
make self-check
make runtime-check
make eval-models
make sync-default-model
make check-default-model
make check

CLI examples:

aman doctor --config ~/.config/aman/config.json --json
aman self-check --config ~/.config/aman/config.json --json
aman run --config ~/.config/aman/config.json
aman bench --text "example transcript" --repeat 5 --warmup 1
aman build-heuristic-dataset --input benchmarks/heuristics_dataset.raw.jsonl --output benchmarks/heuristics_dataset.jsonl --json
aman eval-models --dataset benchmarks/cleanup_dataset.jsonl --matrix benchmarks/model_matrix.small_first.json --heuristic-dataset benchmarks/heuristics_dataset.jsonl --heuristic-weight 0.25 --json
aman version
aman init --config ~/.config/aman/config.json --force

Benchmarking

aman bench --text "draft a short email to Marta confirming lunch" --repeat 10 --warmup 2
aman bench --text-file ./bench-input.txt --repeat 20 --json

bench does not capture audio and never injects text to desktop apps. It runs the processing path from input transcript text through alignment/editor/fact-guard/vocabulary cleanup and prints timing summaries.

Model evaluation

aman build-heuristic-dataset --input benchmarks/heuristics_dataset.raw.jsonl --output benchmarks/heuristics_dataset.jsonl
aman eval-models --dataset benchmarks/cleanup_dataset.jsonl --matrix benchmarks/model_matrix.small_first.json --heuristic-dataset benchmarks/heuristics_dataset.jsonl --heuristic-weight 0.25 --output benchmarks/results/latest.json
make sync-default-model

eval-models runs a structured model/parameter sweep over a JSONL dataset and outputs latency plus quality metrics.
When --heuristic-dataset is provided, the report also includes alignment-heuristic quality metrics.
make sync-default-model promotes the report winner to the managed default model constants and make check-default-model keeps that drift check in CI.

Internal maintainer CLI:

aman-maint sync-default-model --check --report benchmarks/results/latest.json --artifacts benchmarks/model_artifacts.json --constants src/constants.py

Dataset and artifact details live in benchmarks/README.md.

3.4 KiB Raw Blame History