pen-stack 6.2.3__tar.gz → 6.2.4__tar.gz

{pen_stack-6.2.3 → pen_stack-6.2.4}/CHANGELOG.md

@@ -3,6 +3,23 @@
 All notable changes to PEN-STACK are documented here. This file follows
 [Keep a Changelog](https://keepachangelog.com/) and the program's phase structure.
 
+## [6.2.4] - 2026-06-12 - v6.2.4: faster grounded narration (LLM backends) (patch)
+
+**Performance + a real fix, from benchmarking the narration backends on the deployment GPU (RTX A4000).** The
+LLM only narrates over engine tool results (the grounding guard runs regardless), so latency is the thing to
+tune. Measured (real grounded prompt): Ollama qwen2.5:**7b** = ~50s warm / 120s cold (3.4 tok/s — a workstation
+GPU is slow for LLM); qwen2.5:**3b** = ~27s; llama3.2:3b = ~17s; hosted **Nemotron‑49B = ~5s**. All keep the
+grounding guard clean (0 ungrounded). The old Nemotron fallback model (`llama-3.1-nemotron-70b-instruct`) now
+**404s** — i.e. the fallback was silently broken.
+
+### Fixed / changed (`pen_stack/web/llm.py`)
+- **Nemotron model corrected** to `nvidia/llama-3.3-nemotron-super-49b-v1` (the 70B name was retired → the
+  fallback 404'd). Restores a working, ~5s hosted backend.
+- **Default local model qwen2.5:7b → qwen2.5:3b‑instruct** (~2× faster narration, grounding unchanged).
+- **`keep_alive` (default 30m)** so the model stays resident on the GPU and idle calls don't pay the cold start.
+- **Configurable backend order** `PEN_STACK_LLM_ORDER` (default `ollama,nemotron` = local/private‑first;
+  `nemotron,ollama` = speed‑first ~5s). `web/.env.example` documents all knobs.
+
 ## [6.2.3] - 2026-06-12 - v6.2.3: scope matcher covers functional titer + durability (patch)
 
 **Honesty-coverage fix (from the 20-query acceptance suite, `phase_6.2/PEN-STACK_ACCEPTANCE_TESTS.md`).** The

{pen_stack-6.2.3 → pen_stack-6.2.4}/CITATION.cff

@@ -1,7 +1,7 @@
 cff-version: 1.2.0
 message: "If you use PEN-STACK, please cite it as below."
 title: "PEN-STACK: open infrastructure for genome writing"
-version: 6.2.3
+version: 6.2.4
 date-released: 2026-06-11
 authors:
   - family-names: "Mahaboob Ali"

{pen_stack-6.2.3 → pen_stack-6.2.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pen-stack
-Version: 6.2.3
+Version: 6.2.4
 Summary: Open infrastructure for genome writing: the Writable Genome atlas, the Writer Atlas, and the Write Planner.
 Author-email: Anees Ahmed Mahaboob Ali <ahmedaneesm@gmail.com>
 License: MIT
@@ -90,7 +90,7 @@ every design against rule-grounded mechanism, reports calibrated confidence, cit
 [![codecov](https://codecov.io/gh/ahmedanees-m/pen-stack/branch/main/graph/badge.svg)](https://codecov.io/gh/ahmedanees-m/pen-stack)
 [![License: MIT](https://img.shields.io/badge/License-MIT-informational.svg)](LICENSE)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue.svg)](https://www.python.org/)
-[![Version](https://img.shields.io/badge/version-6.2.3-blue.svg)](CHANGELOG.md)
+[![Version](https://img.shields.io/badge/version-6.2.4-blue.svg)](CHANGELOG.md)
 [![Status](https://img.shields.io/badge/status-1.0%20First%20Stable-success.svg)](docs/STABILITY.md)
 [![Tests](https://img.shields.io/badge/tests-371%20passing-success.svg)](tests/)
 [![Lint: ruff](https://img.shields.io/badge/lint-ruff-purple.svg)](https://github.com/astral-sh/ruff)

{pen_stack-6.2.3 → pen_stack-6.2.4}/README.md

@@ -15,7 +15,7 @@ every design against rule-grounded mechanism, reports calibrated confidence, cit
 [![codecov](https://codecov.io/gh/ahmedanees-m/pen-stack/branch/main/graph/badge.svg)](https://codecov.io/gh/ahmedanees-m/pen-stack)
 [![License: MIT](https://img.shields.io/badge/License-MIT-informational.svg)](LICENSE)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue.svg)](https://www.python.org/)
-[![Version](https://img.shields.io/badge/version-6.2.3-blue.svg)](CHANGELOG.md)
+[![Version](https://img.shields.io/badge/version-6.2.4-blue.svg)](CHANGELOG.md)
 [![Status](https://img.shields.io/badge/status-1.0%20First%20Stable-success.svg)](docs/STABILITY.md)
 [![Tests](https://img.shields.io/badge/tests-371%20passing-success.svg)](tests/)
 [![Lint: ruff](https://img.shields.io/badge/lint-ruff-purple.svg)](https://github.com/astral-sh/ruff)

{pen_stack-6.2.3 → pen_stack-6.2.4}/pen_stack/__init__.py

@@ -1,2 +1,2 @@
 """PEN-STACK v3.0 - open infrastructure for genome writing."""
-__version__ = "6.2.3"
+__version__ = "6.2.4"

{pen_stack-6.2.3 → pen_stack-6.2.4}/pen_stack/web/llm.py

@@ -147,10 +147,11 @@ def _call_ollama(prompt: str) -> str:
 
     r = requests.post(
         f"{_ollama_base()}/api/generate",
-        json={"model": os.getenv("OLLAMA_MODEL", "qwen2.5:7b-instruct"),
+        json={"model": os.getenv("OLLAMA_MODEL", "qwen2.5:3b-instruct"),       # 3B narrates ~2x faster than 7B
               "prompt": prompt, "system": SYSTEM, "stream": False,
+              "keep_alive": os.getenv("OLLAMA_KEEP_ALIVE", "30m"),             # keep warm -> avoid the cold start
               "options": {"temperature": 0.2,
-                          "num_predict": int(os.getenv("OLLAMA_NUM_PREDICT", "450"))}},
+                          "num_predict": int(os.getenv("OLLAMA_NUM_PREDICT", "400"))}},
         timeout=_llm_timeout())
     r.raise_for_status()
     return r.json()["response"]
@@ -174,9 +175,9 @@ def _call_nemotron(prompt: str) -> str:
     r = requests.post(
         "https://integrate.api.nvidia.com/v1/chat/completions",
         headers={"Authorization": f"Bearer {key}"},
-        json={"model": os.getenv("NEMOTRON_MODEL", "nvidia/llama-3.1-nemotron-70b-instruct"),
+        json={"model": os.getenv("NEMOTRON_MODEL", "nvidia/llama-3.3-nemotron-super-49b-v1"),
               "messages": [{"role": "system", "content": SYSTEM}, {"role": "user", "content": prompt}],
-              "temperature": 0.2, "max_tokens": 900},
+              "temperature": 0.2, "max_tokens": 600},
         timeout=_llm_timeout())
     r.raise_for_status()
     return r.json()["choices"][0]["message"]["content"]
@@ -207,11 +208,17 @@ def grounded_reply(message: str, history: list | None = None, *, allow_llm: bool
     prompt = _prompt(message, tool_results, history)
 
     if allow_llm and os.getenv("PEN_STACK_NO_LLM") != "1":
-        for backend in (_call_ollama, _call_nemotron):
+        # configurable order: "ollama,nemotron" (privacy/local-first, the default) or "nemotron,ollama"
+        # (speed-first — the hosted 49B answers in ~5s vs ~25-50s on a workstation GPU). Either way the guard runs.
+        backends = {"ollama": _call_ollama, "nemotron": _call_nemotron}
+        order = [b.strip().lower() for b in os.getenv("PEN_STACK_LLM_ORDER", "ollama,nemotron").split(",")]
+        for name in order:
+            fn = backends.get(name)
+            if fn is None:
+                continue
             try:
-                text = _enforce_grounding(backend(prompt), grounded)   # HARD GATE
-                return {"reply": text, "tool_results": tool_results, "grounded": True,
-                        "backend": backend.__name__.removeprefix("_call_")}
+                text = _enforce_grounding(fn(prompt), grounded)        # HARD GATE
+                return {"reply": text, "tool_results": tool_results, "grounded": True, "backend": name}
             except Exception:                                  # any backend failure → next, then deterministic
                 continue
 

{pen_stack-6.2.3 → pen_stack-6.2.4}/pen_stack.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: pen-stack
-Version: 6.2.3
+Version: 6.2.4
 Summary: Open infrastructure for genome writing: the Writable Genome atlas, the Writer Atlas, and the Write Planner.
 Author-email: Anees Ahmed Mahaboob Ali <ahmedaneesm@gmail.com>
 License: MIT
@@ -90,7 +90,7 @@ every design against rule-grounded mechanism, reports calibrated confidence, cit
 [![codecov](https://codecov.io/gh/ahmedanees-m/pen-stack/branch/main/graph/badge.svg)](https://codecov.io/gh/ahmedanees-m/pen-stack)
 [![License: MIT](https://img.shields.io/badge/License-MIT-informational.svg)](LICENSE)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue.svg)](https://www.python.org/)
-[![Version](https://img.shields.io/badge/version-6.2.3-blue.svg)](CHANGELOG.md)
+[![Version](https://img.shields.io/badge/version-6.2.4-blue.svg)](CHANGELOG.md)
 [![Status](https://img.shields.io/badge/status-1.0%20First%20Stable-success.svg)](docs/STABILITY.md)
 [![Tests](https://img.shields.io/badge/tests-371%20passing-success.svg)](tests/)
 [![Lint: ruff](https://img.shields.io/badge/lint-ruff-purple.svg)](https://github.com/astral-sh/ruff)

{pen_stack-6.2.3 → pen_stack-6.2.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "pen-stack"
-version = "6.2.3"
+version = "6.2.4"
 description = "Open infrastructure for genome writing: the Writable Genome atlas, the Writer Atlas, and the Write Planner."
 readme = "README.md"
 requires-python = ">=3.11"