Compare commits

...

2 Commits

Author SHA1 Message Date
kimi
9578330c87 fix: replace wildcard CORS default with explicit localhost origins
All checks were successful
Tests / lint (pull_request) Successful in 4s
Tests / test (pull_request) Successful in 1m14s
The cors_origins setting defaulted to ["*"], which passed through
unchanged in production (non-debug) mode. Now defaults to explicit
localhost origins, and _get_cors_origins() strips any wildcards in
production with a warning.

Fixes #462

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-19 14:58:10 -04:00
92594ea588 [loop-cycle] feat: implement source distinction in system prompts (#463) (#464)
All checks were successful
Tests / lint (push) Successful in 3s
Tests / test (push) Successful in 1m28s
2026-03-19 14:49:31 -04:00
4 changed files with 38 additions and 10 deletions

View File

@@ -138,7 +138,12 @@ 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] = ["*"]
cors_origins: list[str] = [
"http://localhost:3000",
"http://localhost:8000",
"http://127.0.0.1:3000",
"http://127.0.0.1:8000",
]
# Trusted hosts for the Host header check (TrustedHostMiddleware).
# Set TRUSTED_HOSTS as a comma-separated list. Wildcards supported (e.g. "*.ts.net").

View File

@@ -484,15 +484,19 @@ app = FastAPI(
def _get_cors_origins() -> list[str]:
"""Get CORS origins from settings, with sensible defaults."""
"""Get CORS origins from settings, rejecting wildcards in production."""
origins = settings.cors_origins
if settings.debug and origins == ["*"]:
return [
"http://localhost:3000",
"http://localhost:8000",
"http://127.0.0.1:3000",
"http://127.0.0.1:8000",
]
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",
]
return origins

View File

@@ -300,7 +300,11 @@ 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## Memory Context\n\n{memory_context}"
full_prompt = (
f"{base_prompt}\n\n"
f"## GROUNDED CONTEXT (verified sources — cite when using)\n\n"
f"{memory_context}"
)
else:
full_prompt = base_prompt
except Exception as exc:

View File

@@ -23,6 +23,9 @@ 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:
@@ -78,6 +81,18 @@ 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)