Spaces:
Sleeping
Sleeping
Commit ·
0e7bbcc
1
Parent(s): e67c48b
Fix RAGForge v1.5.2 citation helper runtime import
Browse files- README.md +7 -9
- docs/MIGRATION_1.5.2.md +29 -0
- pyproject.toml +1 -1
- src/ragforge/__init__.py +1 -1
- src/ragforge/api.py +1 -1
- src/ragforge/pipeline.py +1 -0
- src/ragforge/ui.py +2 -2
- tests/test_evaluation_assets.py +7 -1
- tests/test_pipeline_helpers.py +16 -0
- tests/test_ui_copy.py +1 -1
README.md
CHANGED
|
@@ -10,21 +10,19 @@ pinned: false
|
|
| 10 |
|
| 11 |
# RAGForge
|
| 12 |
|
| 13 |
-
**RAGForge v1.5.
|
| 14 |
|
| 15 |
RAGForge combines hybrid document retrieval, source-level/hierarchical retrieval, semantic query planning, corrective RAG, Self-RAG-style verification, Text2SQL and an “Ask-the-Web” research path in one CPU-friendly application. The default LLM is **Google Gemini 3.5 Flash-Lite**; the UI also exposes Gemini 3.1 Flash-Lite and stronger Flash models.
|
| 16 |
|
| 17 |
|
| 18 |
-
## What is new in v1.5.
|
| 19 |
|
| 20 |
-
v1.5.
|
| 21 |
|
| 22 |
-
- **
|
| 23 |
-
- **
|
| 24 |
-
- **
|
| 25 |
-
- **
|
| 26 |
-
- **Citation-tail cleanup** - repeated trailing citation sets are deduplicated to reduce output such as `[D3, D6]. [D6] [D3]`.
|
| 27 |
-
- **v1.5 evaluation/retrieval improvements retained** - typed Text2SQL checks, adaptive reranking, saved evaluation history and incremental Deep reuse remain unchanged.
|
| 28 |
|
| 29 |
## What is new in v1.5
|
| 30 |
|
|
|
|
| 10 |
|
| 11 |
# RAGForge
|
| 12 |
|
| 13 |
+
**RAGForge v1.5.2 - a production-style, portfolio-ready agentic RAG system for Hugging Face Spaces.**
|
| 14 |
|
| 15 |
RAGForge combines hybrid document retrieval, source-level/hierarchical retrieval, semantic query planning, corrective RAG, Self-RAG-style verification, Text2SQL and an “Ask-the-Web” research path in one CPU-friendly application. The default LLM is **Google Gemini 3.5 Flash-Lite**; the UI also exposes Gemini 3.1 Flash-Lite and stronger Flash models.
|
| 16 |
|
| 17 |
|
| 18 |
+
## What is new in v1.5.2
|
| 19 |
|
| 20 |
+
v1.5.2 is a narrow runtime hotfix over v1.5.1. It does not change retrieval, evaluation semantics, dependency pins or the demo benchmark.
|
| 21 |
|
| 22 |
+
- **Quick/Standard/Deep evaluation crash fixed** - `pipeline.py` now explicitly imports the citation normalization and repair helpers introduced in v1.5.1. This fixes `NameError: name 'repair_missing_citations' is not defined` during generated-answer evaluation.
|
| 23 |
+
- **Packaging regression coverage** - a dependency-free test now verifies the citation helpers are imported into the pipeline namespace, and the LangGraph-backed helper test invokes the repair path directly when optional runtime dependencies are available.
|
| 24 |
+
- **v1.5.1 behavior retained** - saved-report JSON, table export, grouped-citation parsing and conservative citation repair are unchanged.
|
| 25 |
+
- **Benchmark compatibility retained** - the benchmark remains `1.5.1` because this hotfix changes runtime wiring only, not benchmark questions or scoring semantics.
|
|
|
|
|
|
|
| 26 |
|
| 27 |
## What is new in v1.5
|
| 28 |
|
docs/MIGRATION_1.5.2.md
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Migration to RAGForge v1.5.2
|
| 2 |
+
|
| 3 |
+
v1.5.2 is a narrow hotfix over v1.5.1.
|
| 4 |
+
|
| 5 |
+
## Fixed
|
| 6 |
+
|
| 7 |
+
Quick, Standard and Deep evaluation could fail while processing generated answers with:
|
| 8 |
+
|
| 9 |
+
```text
|
| 10 |
+
NameError: name 'repair_missing_citations' is not defined
|
| 11 |
+
```
|
| 12 |
+
|
| 13 |
+
`src/ragforge/pipeline.py` called citation helpers moved into `src/ragforge/citations.py`, but the v1.5.1 package omitted the explicit module import. v1.5.2 restores:
|
| 14 |
+
|
| 15 |
+
```python
|
| 16 |
+
from .citations import normalize_citation_syntax, repair_missing_citations
|
| 17 |
+
```
|
| 18 |
+
|
| 19 |
+
## Compatibility
|
| 20 |
+
|
| 21 |
+
- No dependency changes.
|
| 22 |
+
- No corpus/index format changes.
|
| 23 |
+
- No retrieval-policy changes.
|
| 24 |
+
- No evaluation-scoring changes.
|
| 25 |
+
- The demo benchmark remains version `1.5.1`, so saved reports are semantically compatible while the same server-side workspace exists.
|
| 26 |
+
|
| 27 |
+
## Regression coverage
|
| 28 |
+
|
| 29 |
+
The test suite now checks the import wiring without optional dependencies and invokes the citation-repair pipeline helper when LangGraph is available.
|
pyproject.toml
CHANGED
|
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
|
|
| 4 |
|
| 5 |
[project]
|
| 6 |
name = "ragforge"
|
| 7 |
-
version = "1.5.
|
| 8 |
description = "Production-style agentic RAG demo for Hugging Face Spaces"
|
| 9 |
requires-python = ">=3.11"
|
| 10 |
dependencies = []
|
|
|
|
| 4 |
|
| 5 |
[project]
|
| 6 |
name = "ragforge"
|
| 7 |
+
version = "1.5.2"
|
| 8 |
description = "Production-style agentic RAG demo for Hugging Face Spaces"
|
| 9 |
requires-python = ">=3.11"
|
| 10 |
dependencies = []
|
src/ragforge/__init__.py
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
"""RAGForge: production-style agentic retrieval augmented generation demo."""
|
| 2 |
|
| 3 |
-
__version__ = "1.5.
|
|
|
|
| 1 |
"""RAGForge: production-style agentic retrieval augmented generation demo."""
|
| 2 |
|
| 3 |
+
__version__ = "1.5.2"
|
src/ragforge/api.py
CHANGED
|
@@ -29,7 +29,7 @@ def _auth(authorization: Annotated[str | None, Header()] = None) -> None:
|
|
| 29 |
|
| 30 |
|
| 31 |
def create_api() -> FastAPI:
|
| 32 |
-
app = FastAPI(title="RAGForge API", version="1.5.
|
| 33 |
|
| 34 |
@app.get("/api/health")
|
| 35 |
def health():
|
|
|
|
| 29 |
|
| 30 |
|
| 31 |
def create_api() -> FastAPI:
|
| 32 |
+
app = FastAPI(title="RAGForge API", version="1.5.2")
|
| 33 |
|
| 34 |
@app.get("/api/health")
|
| 35 |
def health():
|
src/ragforge/pipeline.py
CHANGED
|
@@ -11,6 +11,7 @@ from typing import Any, Callable, TypedDict
|
|
| 11 |
from cachetools import TTLCache
|
| 12 |
from langgraph.graph import END, START, StateGraph
|
| 13 |
|
|
|
|
| 14 |
from .config import get_settings
|
| 15 |
from .llm import GeminiGateway, RequestPacer
|
| 16 |
from .schemas import EvidenceAssessment, PipelineConfig, QueryPlan, QueryResponse, SearchHit
|
|
|
|
| 11 |
from cachetools import TTLCache
|
| 12 |
from langgraph.graph import END, START, StateGraph
|
| 13 |
|
| 14 |
+
from .citations import normalize_citation_syntax, repair_missing_citations
|
| 15 |
from .config import get_settings
|
| 16 |
from .llm import GeminiGateway, RequestPacer
|
| 17 |
from .schemas import EvidenceAssessment, PipelineConfig, QueryPlan, QueryResponse, SearchHit
|
src/ragforge/ui.py
CHANGED
|
@@ -243,7 +243,7 @@ def _architecture_snapshot(session_id: str | None) -> tuple[str, str, str, dict[
|
|
| 243 |
settings = get_settings()
|
| 244 |
stats = ws.stats()
|
| 245 |
runtime_json = {
|
| 246 |
-
"ragforge_version": "1.5.
|
| 247 |
"workspace": stats,
|
| 248 |
"models": {
|
| 249 |
"generation": settings.default_model,
|
|
@@ -264,7 +264,7 @@ def _architecture_snapshot(session_id: str | None) -> tuple[str, str, str, dict[
|
|
| 264 |
}
|
| 265 |
runtime = (
|
| 266 |
"### Live runtime\n"
|
| 267 |
-
f"**RAGForge:** `v1.5.
|
| 268 |
f"**Corpus:** `{stats['sources']}` sources - `{stats['chunks']}` chunks - "
|
| 269 |
f"`{stats['source_profiles']}` source profiles - `{stats['tables']}` tables - "
|
| 270 |
f"corpus version `{stats['version']}`\n\n"
|
|
|
|
| 243 |
settings = get_settings()
|
| 244 |
stats = ws.stats()
|
| 245 |
runtime_json = {
|
| 246 |
+
"ragforge_version": "1.5.2",
|
| 247 |
"workspace": stats,
|
| 248 |
"models": {
|
| 249 |
"generation": settings.default_model,
|
|
|
|
| 264 |
}
|
| 265 |
runtime = (
|
| 266 |
"### Live runtime\n"
|
| 267 |
+
f"**RAGForge:** `v1.5.2` - **workspace:** `{sid[:12]}...` - **status:** `{stats['status']}`\n\n"
|
| 268 |
f"**Corpus:** `{stats['sources']}` sources - `{stats['chunks']}` chunks - "
|
| 269 |
f"`{stats['source_profiles']}` source profiles - `{stats['tables']}` tables - "
|
| 270 |
f"corpus version `{stats['version']}`\n\n"
|
tests/test_evaluation_assets.py
CHANGED
|
@@ -49,7 +49,7 @@ def test_demo_evaluation_and_introspection_are_available_through_api():
|
|
| 49 |
assert "/api/v1/session/{session_id}" in text
|
| 50 |
assert "/api/v1/evaluation/saved/{session_id}" in text
|
| 51 |
assert "/api/v1/evaluation/saved/{session_id}/{level}" in text
|
| 52 |
-
assert 'version="1.5.
|
| 53 |
|
| 54 |
|
| 55 |
def test_v15_evaluation_cache_and_incremental_deep_are_present():
|
|
@@ -110,3 +110,9 @@ def test_v151_citation_repair_skips_preamble_colons_and_normalizes_groups():
|
|
| 110 |
assert "_normalize_citation_syntax" in pipeline_text
|
| 111 |
assert 'stripped.endswith(":")' in citation_text
|
| 112 |
assert "group_re" in citation_text
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
assert "/api/v1/session/{session_id}" in text
|
| 50 |
assert "/api/v1/evaluation/saved/{session_id}" in text
|
| 51 |
assert "/api/v1/evaluation/saved/{session_id}/{level}" in text
|
| 52 |
+
assert 'version="1.5.2"' in text
|
| 53 |
|
| 54 |
|
| 55 |
def test_v15_evaluation_cache_and_incremental_deep_are_present():
|
|
|
|
| 110 |
assert "_normalize_citation_syntax" in pipeline_text
|
| 111 |
assert 'stripped.endswith(":")' in citation_text
|
| 112 |
assert "group_re" in citation_text
|
| 113 |
+
|
| 114 |
+
|
| 115 |
+
def test_v152_pipeline_imports_citation_helpers_into_runtime_namespace():
|
| 116 |
+
text = Path("src/ragforge/pipeline.py").read_text(encoding="utf-8")
|
| 117 |
+
assert "from .citations import normalize_citation_syntax, repair_missing_citations" in text
|
| 118 |
+
assert "return repair_missing_citations(answer, sources)" in text
|
tests/test_pipeline_helpers.py
CHANGED
|
@@ -17,3 +17,19 @@ def test_evidence_strength_does_not_treat_rrf_rank_as_calibrated_relevance():
|
|
| 17 |
# A top RRF result may have score=1.0 purely because it ranked first.
|
| 18 |
weak = SearchHit(Chunk("1", "unrelated", "file.md"), score=1.0, dense_score=0.08, sparse_score=0.0)
|
| 19 |
assert RAGEngine._evidence_strength([weak], []) < 0.30
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
# A top RRF result may have score=1.0 purely because it ranked first.
|
| 18 |
weak = SearchHit(Chunk("1", "unrelated", "file.md"), score=1.0, dense_score=0.08, sparse_score=0.0)
|
| 19 |
assert RAGEngine._evidence_strength([weak], []) < 0.30
|
| 20 |
+
|
| 21 |
+
|
| 22 |
+
def test_pipeline_citation_repair_helper_is_runtime_bound():
|
| 23 |
+
sources = [
|
| 24 |
+
{
|
| 25 |
+
"id": "D1",
|
| 26 |
+
"title": "policy.txt",
|
| 27 |
+
"snippet": "Refunds usually appear within 5 to 10 business days after confirmation.",
|
| 28 |
+
}
|
| 29 |
+
]
|
| 30 |
+
repaired, count = RAGEngine._repair_missing_citations(
|
| 31 |
+
"Refunds usually appear within 5 to 10 business days after confirmation.",
|
| 32 |
+
sources,
|
| 33 |
+
)
|
| 34 |
+
assert count >= 1
|
| 35 |
+
assert "[D1]" in repaired
|
tests/test_ui_copy.py
CHANGED
|
@@ -54,7 +54,7 @@ def test_ui_has_quota_safe_evaluation_controls_and_score_card_spacing():
|
|
| 54 |
|
| 55 |
def test_architecture_snapshot_returns_complete_runtime_payload():
|
| 56 |
text = Path("src/ragforge/ui.py").read_text(encoding="utf-8")
|
| 57 |
-
assert '"ragforge_version": "1.5.
|
| 58 |
assert "return sid, runtime, curl, runtime_json" in text
|
| 59 |
assert "curl = f\ndef _eval_frame" not in text
|
| 60 |
|
|
|
|
| 54 |
|
| 55 |
def test_architecture_snapshot_returns_complete_runtime_payload():
|
| 56 |
text = Path("src/ragforge/ui.py").read_text(encoding="utf-8")
|
| 57 |
+
assert '"ragforge_version": "1.5.2"' in text
|
| 58 |
assert "return sid, runtime, curl, runtime_json" in text
|
| 59 |
assert "curl = f\ndef _eval_frame" not in text
|
| 60 |
|