Skip to content

fix(module-b): replace retired noise-filter model default with gemini-3.5-flash-lite #1151

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Pa04rth merged 4 commits into from
Nov 6, 2026
Merged

fix(module-b): replace retired noise-filter model default with gemini-3.5-flash-lite #1151

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
b53a15b
fix(module-b): bump retired noise-filter model default to gemini-2.5-…
skypank-coder Oct 14, 2026
c38d692
Merge branch 'main' into fix/noise-filter-model-default
skypank-coder Oct 14, 2026
262ad54
fix(module-b): replace retired noise-filter model default with gemini…
skypank-coder Oct 14, 2026
10fffef
Merge branch 'main' into fix/noise-filter-model-default
skypank-coder Nov 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .env.example
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,7 +61,7 @@ VERTEX_EMBED_CONTENT_MODEL=gemini-embedding-001

# Noise / Relevance Filter (Module B)

CRE_NOISE_FILTER_LLM_MODEL=gemini/gemini-2.5-flash-lite
CRE_NOISE_FILTER_LLM_MODEL=gemini/gemini-3.5-flash-lite
CRE_NOISE_FILTER_BATCH_SIZE=10
CRE_NOISE_FILTER_MAX_CHARS=1500
CRE_NOISE_FILTER_CONFIDENCE_THRESHOLD=0.8
Expand Down
2 changes: 1 addition & 1 deletion application/utils/noise_filter/config_loader.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@
# Cheap, dedicated classification model. Decoupled from CRE_LLM_CHAT_MODEL by
# design: Module B is the cheap gate and must not pull in the chatbot's
# heavier model.
DEFAULT_LLM_MODEL = "gemini/gemini-2.5-flash-lite"
DEFAULT_LLM_MODEL = "gemini/gemini-3.5-flash-lite"
DEFAULT_BATCH_SIZE = 10
DEFAULT_MAX_CHARS = 1500
DEFAULT_CONFIDENCE_THRESHOLD = 0.8
Expand Down
2 changes: 1 addition & 1 deletion application/utils/noise_filter/llm_classifier.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@
construct PromptHandler (DB-coupled). Completions go through
``application.prompt_client.litellm_router`` so retry policy and parsing match
chat/embeddings. The classifier uses a dedicated cheap model
(config.llm_model, default gemini/gemini-2.5-flash-lite) and never falls back
(config.llm_model, default gemini/gemini-3.5-flash-lite) and never falls back
to CRE_LLM_CHAT_MODEL: Module B is the cheap gate and must stay decoupled
from the chatbot's model.
"""
Expand Down
2 changes: 1 addition & 1 deletion docs/gsoc_2026_module_b/module_b_runbook.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ vector;`).
| Var | Purpose | Default |
|---|---|---|
| `GEMINI_API_KEY` | LLM credential (Gemini) | — (required) |
| `CRE_NOISE_FILTER_LLM_MODEL` | classification model | `gemini/gemini-2.5-flash-lite` |
| `CRE_NOISE_FILTER_LLM_MODEL` | classification model | `gemini/gemini-3.5-flash-lite` |
Comment thread
Pa04rth marked this conversation as resolved.
Resolved
| `CRE_NOISE_FILTER_BATCH_SIZE` | chunks per LLM request | `10` |
| `CRE_NOISE_FILTER_MAX_CHARS` | per-chunk truncation | `1500` |
| `CRE_NOISE_FILTER_CONFIDENCE_THRESHOLD` | KNOWLEDGE-below → UNCERTAIN | `0.8` |
Expand Down
Loading