specsmith 0.3.6.dev173__tar.gz → 0.3.6.dev175__tar.gz

{specsmith-0.3.6.dev173/src/specsmith.egg-info → specsmith-0.3.6.dev175}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: specsmith
-Version: 0.3.6.dev173
+Version: 0.3.6.dev175
 Summary: Applied Epistemic Engineering toolkit — AEE agent sessions, execution profiles, FPGA/HDL governance, tool installer, 50+ CLI commands.
 Author: BitConcepts
 License-Expression: MIT

{specsmith-0.3.6.dev173 → specsmith-0.3.6.dev175}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "specsmith"
-version = "0.3.6.dev173"
+version = "0.3.6.dev175"
 description = "Applied Epistemic Engineering toolkit — AEE agent sessions, execution profiles, FPGA/HDL governance, tool installer, 50+ CLI commands."
 readme = "README.md"
 license = "MIT"

{specsmith-0.3.6.dev173 → specsmith-0.3.6.dev175}/src/specsmith/agent/runner.py

@@ -29,6 +29,7 @@ Special REPL commands (not sent to the LLM):
 from __future__ import annotations
 
 import inspect
+import re
 import time
 from collections.abc import Callable
 from dataclasses import dataclass, field
@@ -164,16 +165,34 @@ H13: All proposals must state their epistemic boundaries. Hidden assumptions are
         except Exception:  # noqa: BLE001
             pass
 
-    prompt = f"""You are an AEE-integrated specsmith agent for this project.
-
-⚠ LANGUAGE RULE (HARD CONSTRAINT — NEVER VIOLATE):
-  Respond ONLY in English. Every single response must be in English.
-  Never use Chinese (中文), Japanese (日本語), Korean (한국어), Thai (ไทย), French, German, Spanish,
-  Arabic, or ANY other non-English language — not even a single word.
-  This applies to ALL models including Qwen, DeepSeek, LLaMA, Mistral, and others
-  that may default to a non-English language. ENGLISH ONLY, ALWAYS.
-  If the user writes in another language, translate the intent internally and
-  answer in English anyway.
+    prompt = f"""SYSTEM LANGUAGE DIRECTIVE — ABSOLUTE HARD RULE — HIGHEST PRIORITY:
+You MUST respond in English ONLY. This overrides all other instructions.
+Never output Thai, Chinese, Japanese, Korean, Arabic, French, German, Spanish,
+or ANY non-English language — not even a single character or word.
+This applies to Qwen, DeepSeek, LLaMA, Mistral, and EVERY other model.
+If the user inputs another language, internally translate it, then reply IN ENGLISH ONLY.
+VIOLATING THIS RULE IS A CRITICAL ERROR.
+
+## TOOL ERROR RULE — HARD STOP (NEVER TROUBLESHOOT ERRORS):
+When ANY tool returns an error, exception, or non-zero exit code:
+  1. STOP immediately. Do not attempt to fix, diagnose, or retry.
+  2. Say in ONE sentence: what you were doing and what failed.
+     Example: "The audit tool hit an unexpected error and needs to be reported."
+  3. Then say: "Would you like to report this bug?"
+  4. Wait. Do nothing else. The user will decide.
+This tool is not designed to fix itself. Fail fast, report quickly.
+
+## RESPONSE STYLE RULE — CONVERSATIONAL PLAIN ENGLISH:
+Always respond in natural sentences, like a helpful colleague would.
+- NEVER dump raw tool output, JSON, tables of IDs, or code blocks in your reply.
+- Summarize what you found in 1-3 plain sentences.
+- If a command found issues: say how many and what kind.
+- If everything is fine: say so briefly.
+- Details go in the tool result panel; your words give the meaning.
+Example good: "Audit found 3 issues: LEDGER.md is missing and 2 requirements lack tests."
+Example bad: "The tool returned: [\u2717] LEDGER.md MISSING, [\u2717] REQ-001 uncovered..."
+
+You are an AEE-integrated specsmith agent for this project.
 
 ## Project Governance
 {governance_text}
@@ -281,6 +300,7 @@ class AgentRunner:
         self._skills: list[Skill] = load_skills(Path(self.project_dir))
         self._hooks = HookRegistry()
         self._system_prompt = ""
+        self._hard_stop: bool = False  # set True when a critical tool crash is detected
 
         # Execution profile — loaded from scaffold.yml at session start
         from specsmith import profiles
@@ -363,8 +383,100 @@ class AgentRunner:
         self._system_prompt = build_system_prompt(self.project_dir, self._skills)
         return self._agent_turn(task, silent=True)
 
+    # Characters in common CJK / Thai / Arabic Unicode blocks
+    _NON_ASCII_BLOCKS = re.compile(
+        r"[\u0600-\u06FF"  # Arabic
+        r"\u0E00-\u0E7F"  # Thai
+        r"\u3000-\u9FFF"  # CJK Unified Ideographs + punctuation + kana
+        r"\uAC00-\uD7AF"  # Korean Hangul
+        r"\uF900-\uFAFF]"  # CJK Compatibility
+    )
+
+    def _has_non_english(self, text: str) -> bool:
+        """Return True if text contains a significant proportion of non-English script."""
+        if not text:
+            return False
+        hits = len(self._NON_ASCII_BLOCKS.findall(text))
+        return hits > 5 and (hits / max(len(text), 1)) > 0.05
+
+    # ---- Critical error patterns that trigger a hard stop ----
+    _CRITICAL_PATTERNS = re.compile(
+        r"Traceback \(most recent call last\)"
+        r"|\[ERROR\]"
+        r"|UnicodeDecodeError"
+        r"|UnicodeEncodeError"
+        r"|ImportError"
+        r"|ModuleNotFoundError"
+        r"|AttributeError: '"
+        r"|TypeError: unsupported"
+        r"|PermissionError"
+        r"|OSError: "
+        r"|RuntimeError: ",
+        re.IGNORECASE,
+    )
+
+    @staticmethod
+    def _is_critical_error(output: str) -> bool:
+        """Return True if tool output indicates an unexpected crash.
+
+        Normal governance failures (audit issues, missing files) are NOT
+        critical — only Python exceptions and import errors are.
+        """
+        if not output:
+            return False
+        # Non-zero exit alone is expected (e.g. audit found issues).
+        # Only flag when a Python exception signature is present.
+        return AgentRunner._CRITICAL_PATTERNS.search(output) is not None
+
+    def _collect_diagnostics(self, tool_name: str, output: str) -> dict:
+        """Collect diagnostic context for a crash report."""
+        import platform as _platform
+        import sys as _sys
+
+        from specsmith import __version__ as _ver
+
+        project_type = ""
+        try:
+            import yaml as _yaml
+
+            sf = Path(self.project_dir) / "scaffold.yml"
+            if sf.exists():
+                raw = _yaml.safe_load(sf.read_text(encoding="utf-8")) or {}
+                project_type = str(raw.get("type", ""))
+        except Exception:  # noqa: BLE001
+            pass
+
+        # Classify repo: Python exceptions from specsmith module → specsmith CLI
+        # Extension/bridge errors would never reach here (they don’t use this runner)
+        repo = "specsmith"
+
+        # Extract first meaningful error line for the summary
+        summary = output.strip().splitlines()
+        _err_pat = re.compile(r"\w+Error|Exception|RuntimeError")
+        summary_line = next(
+            (ln.strip() for ln in reversed(summary) if _err_pat.match(ln.strip())),
+            summary[0] if summary else "Unknown error",
+        )[:200]
+
+        return {
+            "tool": tool_name,
+            "summary": summary_line,
+            "detail": output[:4000],
+            "specsmith_version": _ver,
+            "python_version": _sys.version.split()[0],
+            "os_info": f"{_platform.system()} {_platform.release()}",
+            "project_type": project_type,
+            "repo": repo,
+        }
+
     def _agent_turn(self, user_input: str, silent: bool = False) -> str:
         """Execute one user→agent turn with tool loop."""
+        # Inject a lightweight English-only reminder into every user message.
+        # This is the most reliable way to keep local models (Qwen, DeepSeek) on track
+        # because many fine-tunes treat the instruction prefix as a per-turn directive.
+        _ENG_PFXS = ("[ENGLISH ONLY]", "[RESPOND IN ENGLISH", "[LANG:EN]")
+        if not any(user_input.startswith(p) for p in _ENG_PFXS):
+            user_input = "[LANG:EN] " + user_input
         # Add user message
         self._state.messages.append(Message(role=Role.USER, content=user_input))
 
@@ -403,14 +515,33 @@ class AgentRunner:
                 final_response = response.content
 
             if not response.has_tool_calls:
+                # Non-English correction: if response appears to be in another language,
+                # issue a single correction turn rather than showing the wrong-language response.
+                if response.content and self._has_non_english(response.content) and _iteration == 0:
+                    correction = (
+                        "[LANG:EN] CRITICAL: Your last response was in a non-English language. "
+                        "You MUST respond in English ONLY. Please re-answer in English."
+                    )
+                    self._state.messages.append(
+                        Message(role=Role.ASSISTANT, content=response.content)
+                    )
+                    self._state.messages.append(Message(role=Role.USER, content=correction))
+                    # Continue the loop to get an English response
+                    continue
                 # Final response — add to history
                 self._state.messages.append(Message(role=Role.ASSISTANT, content=response.content))
                 break
 
             # Process tool calls
+            self._hard_stop = False  # reset before each batch
             tool_results = self._execute_tool_calls(response.tool_calls, silent=silent)
             self._state.tool_calls_made += len(tool_results)
 
+            # Fail fast: a critical tool crash was detected — break immediately
+            # without sending the error back to the LLM (which would try to fix it).
+            if self._hard_stop:
+                break
+
             # Add assistant message with tool calls
             self._state.messages.append(
                 Message(
@@ -508,7 +639,12 @@ class AgentRunner:
     def _execute_tool_calls(
         self, tool_calls: list[dict[str, Any]], silent: bool = False
     ) -> list[ToolResult]:
-        """Execute tool calls and return results."""
+        """Execute tool calls and return results.
+
+        Sets ``self._hard_stop = True`` if any tool produces a critical error
+        (Python exception, import error, etc.) so the caller can break the
+        agentic loop immediately without sending the error to the LLM.
+        """
         from specsmith import profiles as _profiles
 
         results: list[ToolResult] = []
@@ -634,6 +770,22 @@ class AgentRunner:
 
             elapsed = time.time() * 1000 - start_ms
 
+            # ---- Fail-fast: detect critical errors -------------------------
+            # A critical error is an unexpected crash (Python exception, import
+            # failure, etc.) — NOT a normal governance failure (audit issues,
+            # missing files) which the LLM should describe conversationally.
+            if self._is_critical_error(output):
+                self._hard_stop = True
+                diagnostics = self._collect_diagnostics(name, output)
+                if not silent and self._json_events:
+                    self._emit_event(type="tool_crash", **diagnostics)
+                elif not silent:
+                    self._print(
+                        f"\n[CRITICAL ERROR in {name}] "
+                        f"{diagnostics['summary']}\n"
+                        "Session stopped. Please report this bug."
+                    )
+
             if not silent:
                 if self._json_events:
                     self._emit_event(type="tool_finished", name=name, result=output, is_error=error)

{specsmith-0.3.6.dev173 → specsmith-0.3.6.dev175}/src/specsmith/agent/tools.py

@@ -47,6 +47,8 @@ def _run_specsmith(args: list[str], project_dir: str = ".") -> str:
             cmd,
             capture_output=True,
             text=True,
+            encoding="utf-8",  # always decode as UTF-8, not the system locale (cp1252 on Windows)
+            errors="replace",  # replace un-decodable bytes rather than raising UnicodeDecodeError
             timeout=120,
             env=_SUBPROCESS_ENV,
         )

{specsmith-0.3.6.dev173 → specsmith-0.3.6.dev175/src/specsmith.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: specsmith
-Version: 0.3.6.dev173
+Version: 0.3.6.dev175
 Summary: Applied Epistemic Engineering toolkit — AEE agent sessions, execution profiles, FPGA/HDL governance, tool installer, 50+ CLI commands.
 Author: BitConcepts
 License-Expression: MIT