Skip to content

feat: topic graph quality scoring and auto-prune pipeline #607

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

Open
scp7 wants to merge 4 commits into
base: main
Choose a base branch
from
Open

feat: topic graph quality scoring and auto-prune pipeline #607

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
59d5670
feat: add topic scoring and threshold optimization
scp7 Feb 21, 2026
7250378
feat: add sibling coherence and 4-step cascading pruning pipeline
scp7 Feb 25, 2026
39cd257
feat: add YAML scoring config, pipeline auto-prune, and Option B file…
scp7 Feb 26, 2026
fbc65e9
docs: add scoring config reference and pipeline integration docs
scp7 Feb 26, 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
502 changes: 499 additions & 3 deletions deepfabric/cli.py
View file
Open in desktop

Large diffs are not rendered by default.

44 changes: 43 additions & 1 deletion deepfabric/config.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,6 +1,6 @@
import warnings

from typing import Literal
from typing import Literal, Self
Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot Feb 26, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

The Self type was added in Python 3.11. To maintain compatibility with Python 3.10 as specified in pyproject.toml, you should import Self from typing_extensions for versions older than 3.11.

Suggested change
from typing import Literal, Self
from typing import Literal
from typing_extensions import Self


import yaml

Expand Down Expand Up @@ -80,6 +80,43 @@ class LLMConfig(BaseModel):
)


class ScoringConfig(BaseModel):
"""Configuration for topic graph scoring and pruning."""

parent_coherence: float = Field(
default=0.25, ge=0.0, le=1.0, description="Flag nodes with parent coherence below this"
)
sibling_coherence_lower: float = Field(
default=0.2,
ge=0.0,
le=1.0,
description="Flag nodes with sibling coherence below this (outliers)",
)
sibling_coherence_upper: float = Field(
default=0.68,
ge=0.0,
le=1.0,
description="Flag nodes with sibling coherence above this (repetitive)",
)
prune: bool = Field(
default=True, description="Whether to prune flagged nodes (false = score and report only)"
)
save_report: bool = Field(default=False, description="Save score report JSON alongside graph")
embedding_model: str = Field(
default="all-MiniLM-L6-v2", description="SentenceTransformer model for missing embeddings"
)
embedding_key: str = Field(default="embedding", description="Metadata key for node embeddings")

@model_validator(mode="after")
def validate_sibling_coherence_ordering(self) -> Self:
if self.sibling_coherence_lower >= self.sibling_coherence_upper:
raise ValueError(
f"sibling_coherence_lower ({self.sibling_coherence_lower}) "
f"must be less than sibling_coherence_upper ({self.sibling_coherence_upper})"
)
return self


class TopicsConfig(BaseModel):
"""Configuration for topic generation (tree or graph mode)."""

Expand Down Expand Up @@ -126,6 +163,11 @@ class TopicsConfig(BaseModel):
default=None, description="Optional LLM configuration overrides for topics"
)

# Optional scoring configuration (graph mode only)
scoring: ScoringConfig | None = Field(
default=None, description="Topic graph scoring and pruning configuration"
)


class ConversationConfig(BaseModel):
"""Configuration for conversation structure in generation."""
Expand Down
Loading