MakazhanAlpamys/Soup v0.68.0

v0.68.0 — Anti-trend Insurance

on GitHub

What's New

v0.68.0 — Anti-trend Insurance. Five commands that hedge Soup against paradigm shifts. If 1M-token contexts kill fine-tuning, soup compile (DSPy + GEPA + TextGrad prompt-program compilation) replaces it. If teams hit prompt-cost walls, soup distill-prompt bridges to small FT. If only Apple Foundation Models win on-device, soup apple-adapter ships the converter+signing surface. If personal-LLM flywheels become the shape, soup local-rl captures thumbs into SQLite and emits DPO pairs.

• soup compile <program.py> --eval <suite> [--optimizer mipro|gepa|textgrad|copro|bootstrap_fewshot] — DSPy / GEPA prompt-program compiler. Closed allowlist of 5 optimisers. Frozen CompilePlan re-validates on construction. CompileResult rejects NaN / ±Inf / bool-score / negative iterations. Live orchestrator deferred → v0.68.1.
• soup distill-prompt --traces <jsonl> --teacher <m> --student <m> --strategy sft|preference|kl — Distill prompt-heavy GPT-5 / Claude traces into a small FT plan. Bridge between prompt-engineering and FT worlds. Composes with v0.70 cross-tokenizer KD. Live runner → v0.68.1.
• soup compile-tools <spec.json|yaml> --eval <jsonl> — TextGrad / GEPA tool-schema optimiser over OpenAPI / MCP / GraphQL specs. Composes with v0.46 soup agent. JSON / YAML / YML extension allowlist + cwd containment + symlink rejection. Live optimiser → v0.68.1.
• soup apple-adapter <dir> --direction hf-to-mlx|mlx-to-hf|hf-to-apple|mlx-to-apple [--sign] — Adapter format conversion + optional v0.60 Merkle-root signing. Extends v0.25 MLX backend. Live converter → v0.68.1.
• soup local-rl init|status|record|harvest|train — Personal-LLM flywheel daemon. LIVE today: SQLite schema (thumbs + interactions tables, 0o600 perms on POSIX) + record_thumb (parameterised SQL, ≤16 KiB caps, allowlist CHECK constraint) + harvest_dpo_pairs (last-writes-win dedup → {prompt, chosen, rejected} JSONL). Smaller-scope cousin of v0.58 soup loop. Nightly DPO/KTO/ORPO scheduler → v0.68.1.
• +204 tests (11021 → 11225) across 6 new files. Review-fix coverage: 0 CRITICAL + 4 HIGH + 10 MEDIUM + 4 LOW. Lint clean.

Install / Upgrade

pip install --upgrade soup-cli
# or
docker pull ghcr.io/makazhanalpamys/soup:v0.68.0

Security

• All 5 utility modules enforce cwd containment + os.lstat + S_ISLNK rejection on every file path (TOCTOU defence — mirrors v0.33.0 #22 / v0.43.0 Part C / v0.66.0 policy).
• Bool-as-int rejection before every numeric isinstance(int) check (matches v0.30.0 Candidate policy).
• Null-byte rejection on every operator-supplied string.
• _validate_db_path exposed as public validate_db_path + added to __all__.
• POSIX symlink-skip test predicate corrected to sys.platform != "win32" (Windows has os.symlink but needs elevation — the previous hasattr(os, "symlink") would have silently skipped the rejection coverage on Windows CI).
• SQLite uses parameterised queries throughout — no string interpolation in record_thumb / harvest_dpo_pairs.
• Atomic writes via shared paths.atomic_write_text (mkstemp + os.replace, TOCTOU-safe).
• Every CLI command panel passes user-controlled fields through rich.markup.escape.
• No shell=True anywhere in v0.68.0 code paths (source-grep regression-guarded).

Known Limitations

1. Live runners deferred to v0.68.1 — run_compile (DSPy/GEPA orchestrator), prepare_distill_dataset, run_tool_compile, convert_apple_adapter, and run_nightly_train all raise NotImplementedError with explicit v0.68.1 markers. Schema accepts every flag; CLI invocations are gated by validators so misconfigured runs fail loudly at config load. Same stub-then-live cadence as v0.50.0 / v0.61.0 / v0.62.0 / v0.67.0.
2. Part E nightly scheduler is the only Part E gap — soup local-rl init/status/record/harvest are LIVE today. Operators can harvest their personal-LLM feedback into dpo_pairs.jsonl and feed it to existing soup train --task dpo. The v0.68.1 deliverable is the systemd/launchd cron glue.
3. compile_cmd module name vs compile CLI command — source-file is commands/compile_cmd.py (not compile.py) because Python's builtin compile() would shadow at from soup_cli.commands import compile. CLI surface is unchanged: soup compile.
4. harvest_dpo_pairs last-writes-win dedup — multiple ups + downs for the same prompt collapse to ONE pair using the most recent up + most recent down (deterministic by ts ASC). Operators wanting Cartesian-product pairs (every up × every down per prompt) should harvest the raw thumbs table directly.
5. Apple FoundationModels adapter format is undocumented territory — v0.68.0 ships the schema + plan + validation surface; v0.68.1's live converter will track whatever Apple publishes between now and that release.
6. compile-tools accepts .json / .yaml / .yml only — operators using .toml or other spec formats must convert upstream.
7. No live soup serve --record-thumbs integration — operators must manually wire their inference loop to call record_thumb. A v0.68.x patch may add soup serve --record-thumbs <db> for automatic capture.

Full changelog

v0.67.0...v0.68.0 — v0.67.0...v0.68.0