specsmith 0.17.1.dev669__tar.gz → 0.17.1.dev684__tar.gz

{specsmith-0.17.1.dev669/src/specsmith.egg-info → specsmith-0.17.1.dev684}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: specsmith
-Version: 0.17.1.dev669
+Version: 0.17.1.dev684
 Summary: AEE governance toolkit for AI-assisted development — session preflight gates, multi-agent dispatch, requirements↔test traceability, ESDB persistence, MCP server, and skills for Warp, Cursor, Claude Code, Copilot, Windsurf, and Aider.
 Author: Layer1Labs Silicon, Inc.
 License-Expression: MIT

{specsmith-0.17.1.dev669 → specsmith-0.17.1.dev684}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "specsmith"
-version = "0.17.1.dev669"
+version = "0.17.1.dev684"
 description = "AEE governance toolkit for AI-assisted development — session preflight gates, multi-agent dispatch, requirements↔test traceability, ESDB persistence, MCP server, and skills for Warp, Cursor, Claude Code, Copilot, Windsurf, and Aider."
 readme = "README.md"
 license = "MIT"

{specsmith-0.17.1.dev669 → specsmith-0.17.1.dev684}/src/specsmith/agent/runner.py

@@ -394,7 +394,9 @@ class AgentRunner:
                     lines.append(f"    {s.icon} {s.name:<10} \u2717 {s.note}")
             lines.append("")
             if active_count == 0:
-                lines.append("  \u26a0  No provider available — commands will return no response.")
+                lines.append(
+                    "  \u26a0  No provider available \u2014 commands will return no response."
+                )
             else:
                 # Show multi-model routing if the router is configured (REQ-389).
                 if self._model_router is not None:
@@ -585,6 +587,28 @@ class AgentRunner:
             self._emit_event(type="system", message=hint)
             return None
 
+        if result is None:
+            # All providers failed — give the user an actionable explanation
+            # rather than silent emptiness.
+            import os as _os
+
+            host = _os.environ.get("OLLAMA_HOST", DEFAULT_OLLAMA_HOST).rstrip("/")
+            if _ollama_alive(host):
+                model = _pick_ollama_model(host)
+                hint = (
+                    f"Ollama is running but returned no response "
+                    f"(model: {model}). "
+                    "Try: ollama run " + model
+                )
+            else:
+                hint = (
+                    "No LLM provider available. Options:\n"
+                    "  • Start Ollama: ollama serve\n"
+                    "  • Set ANTHROPIC_API_KEY / OPENAI_API_KEY / GOOGLE_API_KEY"
+                )
+            self._emit_event(type="system", message=hint)
+            return None
+
         # Aggregate metrics into the session state (C1).
         # ``run_chat`` now reports tokens_in / tokens_out / cost_usd off the
         # provider response (Ollama prompt_eval_count + eval_count, OpenAI

specsmith-0.17.1.dev684/src/specsmith/agent/token_pricing.py

@@ -0,0 +1,191 @@
+# SPDX-License-Identifier: MIT
+# Copyright (c) 2026 Layer1Labs Silicon, Inc. All rights reserved.
+"""specsmith.agent.token_pricing — static per-token cost lookup.
+
+Provides a single function :func:`cost_for_tokens` that returns the estimated
+USD cost for a given (provider, model, input_tokens, output_tokens) tuple.
+
+Design contract
+---------------
+- ``ollama`` provider (and its aliases ``llamacpp``, ``vllm``, ``lmstudio``)
+  always returns **0.0** — local inference is free; this is intentional and
+  correct, not a bug or missing data.
+- Known cloud models (OpenAI, Anthropic, Google) use a static pricing table
+  updated to Q2 2026 list prices.
+- Unknown providers/models fall back to a conservative default rate so
+  ``cost_of_pass`` analysis degrades gracefully rather than crashing.
+
+Pricing source
+--------------
+Q2 2026 provider pricing pages.  All figures are in USD per 1 M tokens.
+Update this table when provider pricing changes (bump the table version comment).
+"""
+
+from __future__ import annotations
+
+from typing import Any
+
+# ---------------------------------------------------------------------------
+# Pricing table — (input_usd_per_1M, output_usd_per_1M)
+# Table version: 2026-Q2
+# ---------------------------------------------------------------------------
+
+#: Providers that are always zero-cost (local inference).
+_FREE_PROVIDERS: frozenset[str] = frozenset({"ollama", "llamacpp", "vllm", "lmstudio", "local"})
+
+#: Static pricing table.  Keys are normalised model name substrings matched
+#: from longest to shortest so that more-specific keys win over generic ones.
+_PRICING_PER_1M: dict[str, tuple[float, float]] = {
+    # ── OpenAI GPT-4 family ────────────────────────────────────────────────
+    "gpt-4o-mini": (0.15, 0.60),
+    "gpt-4o": (2.50, 10.00),
+    "gpt-4.1-nano": (0.10, 0.40),
+    "gpt-4.1-mini": (0.40, 1.60),
+    "gpt-4.1": (2.00, 8.00),
+    "gpt-4-turbo": (10.00, 30.00),
+    # ── OpenAI GPT-5 family ────────────────────────────────────────────────
+    "gpt-5-nano": (0.50, 2.00),
+    "gpt-5-mini": (2.00, 8.00),
+    "gpt-5-pro": (50.00, 200.00),
+    "gpt-5.4-nano": (0.30, 1.20),
+    "gpt-5.4-mini": (1.00, 4.00),
+    "gpt-5.4-pro": (20.00, 80.00),
+    "gpt-5.4": (5.00, 20.00),
+    "gpt-5.5-pro": (15.00, 60.00),
+    "gpt-5.5": (3.00, 12.00),
+    "gpt-5.2-codex": (10.00, 40.00),
+    "gpt-5.2-pro": (40.00, 160.00),
+    "gpt-5.2": (10.00, 40.00),
+    "gpt-5.3-codex": (8.00, 32.00),
+    "gpt-5.1": (12.00, 48.00),
+    "gpt-5-codex": (15.00, 60.00),
+    "gpt-5": (15.00, 60.00),
+    # ── OpenAI reasoning ──────────────────────────────────────────────────
+    "o4-mini": (1.10, 4.40),
+    "o3-mini": (1.10, 4.40),
+    "o3": (10.00, 40.00),
+    "o1": (15.00, 60.00),
+    # ── Anthropic ─────────────────────────────────────────────────────────
+    "claude-haiku-4-5": (0.25, 1.25),
+    "claude-haiku": (0.25, 1.25),
+    "claude-sonnet-4-5": (3.00, 15.00),
+    "claude-sonnet-4-6": (3.00, 15.00),
+    "claude-sonnet": (3.00, 15.00),
+    "claude-opus-4-5": (15.00, 75.00),
+    "claude-opus-4-6": (15.00, 75.00),
+    "claude-opus": (15.00, 75.00),
+    # ── Google ────────────────────────────────────────────────────────────
+    "gemini-3.5-flash": (0.15, 0.60),
+    "gemini-3-flash": (0.35, 1.05),
+    "gemini-flash": (0.075, 0.30),
+    "gemini-3.1-pro": (1.25, 5.00),
+    "gemini-pro": (1.25, 5.00),
+    # ── Conservative fallback ─────────────────────────────────────────────
+    "unknown": (3.00, 15.00),
+}
+
+
+def _lookup_rates(model: str) -> tuple[float, float]:
+    """Return (input_usd_per_1M, output_usd_per_1M) for *model*.
+
+    Tries an exact key match first, then a longest-substring match, and
+    finally falls back to the ``"unknown"`` conservative default.
+    """
+    key = model.strip().lower()
+    if key in _PRICING_PER_1M:
+        return _PRICING_PER_1M[key]
+    # Longest-prefix / substring match so that "gpt-4o-mini-2026-04-09"
+    # still resolves to "gpt-4o-mini".
+    best_match = ""
+    for table_key in _PRICING_PER_1M:
+        if table_key in key and len(table_key) > len(best_match):
+            best_match = table_key
+    if best_match:
+        return _PRICING_PER_1M[best_match]
+    return _PRICING_PER_1M["unknown"]
+
+
+def cost_for_tokens(
+    provider: str,
+    model: str,
+    input_tokens: int,
+    output_tokens: int,
+) -> float:
+    """Return estimated USD cost for a single LLM call.
+
+    Args:
+        provider: Provider name (e.g. ``"ollama"``, ``"openai"``, ``"anthropic"``).
+        model: Model name or slug (e.g. ``"gpt-4o"``, ``"qwen2.5-coder:14b"``).
+        input_tokens: Number of prompt/context tokens consumed.
+        output_tokens: Number of generated/completion tokens.
+
+    Returns:
+        Estimated cost in USD.  Always ``0.0`` for local/free providers
+        (Ollama, llamacpp, vllm, lmstudio, local).  Returns ``0.0`` when
+        both token counts are zero regardless of provider.
+
+    Examples::
+
+        >>> cost_for_tokens("ollama", "qwen2.5-coder:14b", 1000, 500)
+        0.0
+        >>> cost_for_tokens("openai", "gpt-4o-mini", 1_000_000, 0)
+        0.15
+        >>> cost_for_tokens("anthropic", "claude-haiku-4-5", 0, 1_000_000)
+        1.25
+    """
+    if not provider and not model:
+        return 0.0
+    if (provider or "").strip().lower() in _FREE_PROVIDERS:
+        return 0.0
+    if input_tokens == 0 and output_tokens == 0:
+        return 0.0
+    inp_per_m, out_per_m = _lookup_rates(model)
+    return (input_tokens * inp_per_m + output_tokens * out_per_m) / 1_000_000
+
+
+def cost_for_tokens_breakdown(
+    provider: str,
+    model: str,
+    input_tokens: int,
+    output_tokens: int,
+) -> tuple[float, float, float]:
+    """Return ``(input_cost_usd, output_cost_usd, total_cost_usd)``."""
+    if (provider or "").strip().lower() in _FREE_PROVIDERS:
+        return 0.0, 0.0, 0.0
+    inp_per_m, out_per_m = _lookup_rates(model)
+    inp_cost = input_tokens * inp_per_m / 1_000_000
+    out_cost = output_tokens * out_per_m / 1_000_000
+    return inp_cost, out_cost, inp_cost + out_cost
+
+
+def tokens_per_correct_answer(
+    records: list[Any],
+    *,
+    token_attr: str = "tokens_total",
+    passed_attr: str = "passed",
+) -> float | None:
+    """Compute mean tokens consumed per correct (passing) answer.
+
+    Args:
+        records: Sequence of record objects (e.g. :class:`MetricsRecord` or
+            :class:`RunResult`) with numeric ``token_attr`` and bool
+            ``passed_attr`` attributes.
+        token_attr: Attribute name for total token count.
+        passed_attr: Attribute name for pass/fail boolean.
+
+    Returns:
+        Mean total tokens among passing records, or ``None`` if no passing
+        records exist.
+    """
+    passing = [r for r in records if getattr(r, passed_attr, False)]
+    if not passing:
+        return None
+    totals = [getattr(r, token_attr, 0) for r in passing]
+    return sum(totals) / len(totals)
+
+
+__all__ = [
+    "cost_for_tokens",
+    "cost_for_tokens_breakdown",
+    "tokens_per_correct_answer",
+]

{specsmith-0.17.1.dev669 → specsmith-0.17.1.dev684}/src/specsmith/auditor.py

@@ -554,8 +554,9 @@ def check_ledger_health(root: Path) -> list[AuditResult]:
 
 
 # Default thresholds (used when no project type is detected)
+# AGENTS.md: raised to 250 (template now generates ~202 lines with Parallel Agents section)
 _DEFAULT_THRESHOLDS: dict[str, int] = {
-    "AGENTS.md": 200,
+    "AGENTS.md": 250,
     "docs/governance/RULES.md": 800,
     "docs/governance/SESSION-PROTOCOL.md": 400,
     "docs/governance/LIFECYCLE.md": 200,

{specsmith-0.17.1.dev669 → specsmith-0.17.1.dev684}/src/specsmith/compliance/_compat.py

@@ -8,6 +8,7 @@ traceability matrix. Used by both CLI commands and REST endpoints.
 
 from __future__ import annotations
 
+import json
 import re
 from dataclasses import dataclass, field
 from pathlib import Path
@@ -41,63 +42,107 @@ class ComplianceSummary:
 
 
 def _parse_requirements(root: Path) -> dict[str, dict[str, str]]:
-    """Parse REQUIREMENTS.md into {req_id: {title, status, ...}}."""
+    """Parse requirements into {req_id: {title, status, ...}}.
+
+    Sources tried in order:
+    1. ``docs/REQUIREMENTS.md`` (legacy markdown mode)
+    2. ``REQUIREMENTS.md`` (legacy markdown mode, repo-root variant)
+    3. ``.specsmith/requirements.json`` (YAML-first sync cache)
+    """
     req_path = root / "docs" / "REQUIREMENTS.md"
     if not req_path.is_file():
         req_path = root / "REQUIREMENTS.md"
-    if not req_path.is_file():
-        return {}
-
-    text = req_path.read_text(encoding="utf-8", errors="replace")
-    reqs: dict[str, dict[str, str]] = {}
-
-    # Parse markdown sections: ## N. Title\n- **ID:** REQ-NNN
-    blocks = re.split(r"\n## ", text)
-    for block in blocks:
-        id_match = re.search(r"\*\*ID:\*\*\s*(REQ-\d+)", block)
-        if not id_match:
-            # Try alternate format: ## REQ-NNN
-            id_match = re.search(r"^(REQ-\d+)", block.strip())
-        if id_match:
-            req_id = id_match.group(1)
-            title_match = re.search(r"\*\*Title:\*\*\s*(.+)", block)
-            status_match = re.search(r"\*\*Status:\*\*\s*(.+)", block)
-            reqs[req_id] = {
-                "id": req_id,
-                "title": title_match.group(1).strip() if title_match else "",
-                "status": status_match.group(1).strip() if status_match else "draft",
+
+    if req_path.is_file():
+        text = req_path.read_text(encoding="utf-8", errors="replace")
+        reqs: dict[str, dict[str, str]] = {}
+        # Parse markdown sections: ## N. Title\n- **ID:** REQ-NNN
+        blocks = re.split(r"\n## ", text)
+        for block in blocks:
+            id_match = re.search(r"\*\*ID:\*\*\s*(REQ-\d+)", block)
+            if not id_match:
+                # Try alternate format: ## REQ-NNN
+                id_match = re.search(r"^(REQ-\d+)", block.strip())
+            if id_match:
+                req_id = id_match.group(1)
+                title_match = re.search(r"\*\*Title:\*\*\s*(.+)", block)
+                status_match = re.search(r"\*\*Status:\*\*\s*(.+)", block)
+                reqs[req_id] = {
+                    "id": req_id,
+                    "title": title_match.group(1).strip() if title_match else "",
+                    "status": status_match.group(1).strip() if status_match else "draft",
+                }
+        return reqs
+
+    # YAML-first fallback: read from the .specsmith sync cache
+    json_path = root / ".specsmith" / "requirements.json"
+    if json_path.is_file():
+        try:
+            entries = json.loads(json_path.read_text(encoding="utf-8"))
+        except (json.JSONDecodeError, OSError):
+            return {}
+        return {
+            entry["id"]: {
+                "id": entry["id"],
+                "title": entry.get("title", ""),
+                "status": entry.get("status", "draft"),
             }
+            for entry in entries
+            if isinstance(entry, dict) and "id" in entry
+        }
 
-    return reqs
+    return {}
 
 
 def _parse_tests(root: Path) -> dict[str, dict[str, str]]:
-    """Parse TESTS.md into {test_id: {title, requirement_id, ...}}."""
+    """Parse test cases into {test_id: {title, requirement_id, ...}}.
+
+    Sources tried in order:
+    1. ``docs/TESTS.md`` (legacy markdown mode)
+    2. ``TESTS.md`` (legacy markdown mode, repo-root variant)
+    3. ``.specsmith/testcases.json`` (YAML-first sync cache)
+    """
     test_path = root / "docs" / "TESTS.md"
     if not test_path.is_file():
         test_path = root / "TESTS.md"
-    if not test_path.is_file():
-        return {}
-
-    text = test_path.read_text(encoding="utf-8", errors="replace")
-    tests: dict[str, dict[str, str]] = {}
-
-    blocks = re.split(r"\n## ", text)
-    for block in blocks:
-        id_match = re.search(r"\*\*ID:\*\*\s*(TEST-\d+)", block)
-        if not id_match:
-            id_match = re.search(r"^(TEST-\d+)", block.strip())
-        if id_match:
-            test_id = id_match.group(1)
-            req_match = re.search(r"\*\*Requirement(?:\s+ID)?:\*\*\s*(REQ-\d+)", block)
-            title_match = re.search(r"\*\*Title:\*\*\s*(.+)", block)
-            tests[test_id] = {
-                "id": test_id,
-                "title": title_match.group(1).strip() if title_match else "",
-                "requirement_id": req_match.group(1) if req_match else "",
+
+    if test_path.is_file():
+        text = test_path.read_text(encoding="utf-8", errors="replace")
+        tests: dict[str, dict[str, str]] = {}
+        blocks = re.split(r"\n## ", text)
+        for block in blocks:
+            id_match = re.search(r"\*\*ID:\*\*\s*(TEST-\d+)", block)
+            if not id_match:
+                id_match = re.search(r"^(TEST-\d+)", block.strip())
+            if id_match:
+                test_id = id_match.group(1)
+                req_match = re.search(r"\*\*Requirement(?:\s+ID)?:\*\*\s*(REQ-\d+)", block)
+                title_match = re.search(r"\*\*Title:\*\*\s*(.+)", block)
+                tests[test_id] = {
+                    "id": test_id,
+                    "title": title_match.group(1).strip() if title_match else "",
+                    "requirement_id": req_match.group(1) if req_match else "",
+                }
+        return tests
+
+    # YAML-first fallback: read from the .specsmith sync cache
+    json_path = root / ".specsmith" / "testcases.json"
+    if json_path.is_file():
+        try:
+            entries = json.loads(json_path.read_text(encoding="utf-8"))
+        except (json.JSONDecodeError, OSError):
+            return {}
+        return {
+            entry["id"]: {
+                "id": entry["id"],
+                "title": entry.get("title", ""),
+                "requirement_id": entry.get("requirement_id", ""),
             }
+            for entry in entries
+            if isinstance(entry, dict) and "id" in entry
+        }
 
-    return tests
+    return {}
 
 
 def get_compliance_summary(project_dir: str | Path = ".") -> ComplianceSummary:

{specsmith-0.17.1.dev669 → specsmith-0.17.1.dev684}/src/specsmith/migrations/__init__.py

@@ -117,6 +117,7 @@ class _MigrationRegistry:
             m007_yaml_first,
             m008_esdb_full_coverage,
             m009_esdb_first,
+            m010_post_esdb_cleanup,
         )
 
         instances: list[Migration] = [
@@ -129,6 +130,7 @@ class _MigrationRegistry:
             m007_yaml_first.YamlFirstMigration(),
             m008_esdb_full_coverage.EsdbFullCoverageMigration(),
             m009_esdb_first.EsdbFirstMigration(),
+            m010_post_esdb_cleanup.PostEsdbCleanupMigration(),
         ]
         instances.sort(key=lambda m: m.version)
         self._migrations = instances