Merge branch 'main' into kimi/issue-454

Replace direct mutation of the global EMBEDDING_MODEL with
patch.object on _get_embedding_model, eliminating shared state
that caused flakiness under pytest-xdist parallel execution.

Fixes #454

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/config.py

@@ -138,12 +138,7 @@ class Settings(BaseSettings):
 
     # CORS allowed origins for the web chat interface (Gitea Pages, etc.)
     # Set CORS_ORIGINS as a comma-separated list, e.g. "http://localhost:3000,https://example.com"
-    cors_origins: list[str] = [
-        "http://localhost:3000",
-        "http://localhost:8000",
-        "http://127.0.0.1:3000",
-        "http://127.0.0.1:8000",
-    ]
+    cors_origins: list[str] = ["*"]
 
     # Trusted hosts for the Host header check (TrustedHostMiddleware).
     # Set TRUSTED_HOSTS as a comma-separated list. Wildcards supported (e.g. "*.ts.net").

src/dashboard/app.py

@@ -484,19 +484,15 @@ app = FastAPI(
 
 
 def _get_cors_origins() -> list[str]:
-    """Get CORS origins from settings, rejecting wildcards in production."""
+    """Get CORS origins from settings, with sensible defaults."""
     origins = settings.cors_origins
-    if not settings.debug and "*" in origins:
-        logger.warning(
-            "Wildcard '*' in CORS_ORIGINS ignored in production — "
-            "set explicit origins via CORS_ORIGINS env var"
-        )
-        origins = [o for o in origins if o != "*"]
-        if not origins:
-            origins = [
-                "http://localhost:3000",
-                "http://localhost:8000",
-            ]
+    if settings.debug and origins == ["*"]:
+        return [
+            "http://localhost:3000",
+            "http://localhost:8000",
+            "http://127.0.0.1:3000",
+            "http://127.0.0.1:8000",
+        ]
     return origins
 
 

src/timmy/agent.py

@@ -300,11 +300,7 @@ def create_timmy(
             max_context = 2000 if not use_tools else 8000
             if len(memory_context) > max_context:
                 memory_context = memory_context[:max_context] + "\n... [truncated]"
-            full_prompt = (
-                f"{base_prompt}\n\n"
-                f"## GROUNDED CONTEXT (verified sources — cite when using)\n\n"
-                f"{memory_context}"
-            )
+            full_prompt = f"{base_prompt}\n\n## Memory Context\n\n{memory_context}"
         else:
             full_prompt = base_prompt
     except Exception as exc:

src/timmy/prompts.py

@@ -23,9 +23,6 @@ Rules:
 - Remember what the user tells you during the conversation.
 - If you don't know something, say so honestly — never fabricate facts.
 - If a request is ambiguous, ask a brief clarifying question before guessing.
-- SOURCE DISTINCTION: When answering from memory or retrieved context, cite it.
-  When answering from your own training, use hedging: "I think", "I believe".
-  The user must be able to tell grounded claims from pattern-matching.
 - Use the user's name if you know it.
 - When you state a fact, commit to it.
 - NEVER attempt arithmetic in your head. If asked to compute anything, respond:
@@ -81,18 +78,6 @@ HONESTY:
 - Never fabricate tool output. Call the tool and wait.
 - If a tool errors, report the exact error.
 
-SOURCE DISTINCTION (SOUL requirement — non-negotiable):
-- Every claim you make comes from one of two places: a verified source you
-  can point to, or your own pattern-matching. The user must be able to tell
-  which is which.
-- When your response uses information from GROUNDED CONTEXT (memory, retrieved
-  documents, tool output), cite it: "From memory:", "According to [source]:".
-- When you are generating from your training data alone, signal it naturally:
-  "I think", "My understanding is", "I believe" — never false certainty.
-- If the user asks a factual question and you have no grounded source, say so:
-  "I don't have a verified source for this — from my training I think..."
-- Prefer "I don't know" over a confident-sounding guess. Refusal over fabrication.
-
 MEMORY (three tiers):
 - Tier 1: MEMORY.md (hot, always loaded)
 - Tier 2: memory/ vault (structured, append-only, date-stamped)