Spaces:
Sleeping
Sleeping
Upload 20 files
Browse files- report_engine.py +65 -1
- requirements.txt +1 -1
- smoke_test.py +8 -0
- tools.py +49 -3
report_engine.py
CHANGED
|
@@ -956,6 +956,56 @@ def _compact_scenario_comparison(df: pd.DataFrame) -> pd.DataFrame:
|
|
| 956 |
wanted=["Display Scenario","Source Model","Storm","Storm Status","Simulation Status","Runoff Error (%)","Flow Error (%)","Peak Subcatchment Runoff","Maximum Storage Depth","Maximum Storage Volume","Peak Link Flow","Maximum Node Inflow","Maximum Node Flooding","Hydraulic Difference"]
|
| 957 |
return out[[c for c in wanted if c in out.columns]].copy()
|
| 958 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 959 |
def generate_report_package(
|
| 960 |
*,
|
| 961 |
metadata: ReportMetadata,
|
|
@@ -972,6 +1022,7 @@ def generate_report_package(
|
|
| 972 |
scenario_records: list[Mapping[str, Any]] | None = None,
|
| 973 |
scenario_reporting_mode: str = "Base report with scenario comparison",
|
| 974 |
preliminary_review_artifacts: Mapping[str, Any] | None = None,
|
|
|
|
| 975 |
) -> dict[str, bytes | str]:
|
| 976 |
"""Generate editable Word report and ZIP package entirely in memory."""
|
| 977 |
criteria = criteria or ReportCriteria()
|
|
@@ -1339,7 +1390,9 @@ def generate_report_package(
|
|
| 1339 |
"scenario_comparison": scenario_comparison if scenario_comparison is not None else pd.DataFrame(),
|
| 1340 |
},
|
| 1341 |
)
|
| 1342 |
-
docx_buffer = io.BytesIO()
|
|
|
|
|
|
|
| 1343 |
base = _safe_name(metadata.project_name)
|
| 1344 |
|
| 1345 |
zip_buffer = io.BytesIO()
|
|
@@ -1377,6 +1430,17 @@ def generate_report_package(
|
|
| 1377 |
if frame is not None and not frame.empty:
|
| 1378 |
zf.writestr(f"tables/scenarios/{sid}_{key}.csv", frame.to_csv(index=False))
|
| 1379 |
zf.writestr("metadata/llm_report_context.json", json.dumps(llm_context, indent=2, default=str))
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1380 |
zf.writestr("metadata/approved_narrative_sections.json", json.dumps(dict(narrative_sections or {}), indent=2, ensure_ascii=False))
|
| 1381 |
zf.writestr("metadata/project_metadata.json", json.dumps(asdict(metadata), indent=2))
|
| 1382 |
zf.writestr("metadata/report_criteria.json", json.dumps(asdict(criteria), indent=2, default=str))
|
|
|
|
| 956 |
wanted=["Display Scenario","Source Model","Storm","Storm Status","Simulation Status","Runoff Error (%)","Flow Error (%)","Peak Subcatchment Runoff","Maximum Storage Depth","Maximum Storage Volume","Peak Link Flow","Maximum Node Inflow","Maximum Node Flooding","Hydraulic Difference"]
|
| 957 |
return out[[c for c in wanted if c in out.columns]].copy()
|
| 958 |
|
| 959 |
+
def _embed_attached_figures(doc, figures: list[Mapping[str, Any]] | None) -> None:
|
| 960 |
+
"""Insert session-attached figures into the built document.
|
| 961 |
+
|
| 962 |
+
Each figure dict: {"figure_id", "path", "caption", "section", "source"}.
|
| 963 |
+
Figures are inserted at the END of the first level-1 section whose
|
| 964 |
+
heading contains the requested section keyword (case-insensitive), i.e.
|
| 965 |
+
immediately before the next Heading-1 paragraph; unmatched sections fall
|
| 966 |
+
back to the end of the document. Client-supplied figures are labelled as
|
| 967 |
+
such — they are illustrative material attached to the audited report,
|
| 968 |
+
not server-verified outputs.
|
| 969 |
+
"""
|
| 970 |
+
if not figures:
|
| 971 |
+
return
|
| 972 |
+
from docx.shared import Inches
|
| 973 |
+
|
| 974 |
+
headings = [(i, p) for i, p in enumerate(doc.paragraphs)
|
| 975 |
+
if p.style.name.startswith("Heading 1")]
|
| 976 |
+
|
| 977 |
+
def anchor_for(section_kw: str):
|
| 978 |
+
kw = (section_kw or "results").strip().lower()
|
| 979 |
+
for pos, (idx, para) in enumerate(headings):
|
| 980 |
+
if kw in para.text.lower():
|
| 981 |
+
if pos + 1 < len(headings):
|
| 982 |
+
return headings[pos + 1][1] # insert before next H1
|
| 983 |
+
return None # matched last section -> append at end
|
| 984 |
+
return None
|
| 985 |
+
|
| 986 |
+
for n, fig in enumerate(figures, start=1):
|
| 987 |
+
path = str(fig.get("path", ""))
|
| 988 |
+
if not path or not Path(path).exists():
|
| 989 |
+
continue
|
| 990 |
+
caption = str(fig.get("caption") or fig.get("figure_id") or f"Attached figure {n}")
|
| 991 |
+
source = str(fig.get("source") or "session-attached")
|
| 992 |
+
label = f"Figure A{n} - {caption} ({source}; illustrative, not a server-verified output)"
|
| 993 |
+
anchor = anchor_for(str(fig.get("section", "results")))
|
| 994 |
+
try:
|
| 995 |
+
if anchor is not None:
|
| 996 |
+
pic_par = anchor.insert_paragraph_before()
|
| 997 |
+
pic_par.add_run().add_picture(path, width=Inches(6.0))
|
| 998 |
+
cap_par = anchor.insert_paragraph_before(label)
|
| 999 |
+
cap_par.style = doc.styles["Caption"] if "Caption" in [s.name for s in doc.styles] else cap_par.style
|
| 1000 |
+
else:
|
| 1001 |
+
doc.add_paragraph().add_run().add_picture(path, width=Inches(6.0))
|
| 1002 |
+
doc.add_paragraph(label)
|
| 1003 |
+
except Exception:
|
| 1004 |
+
# A corrupt image must never abort report generation.
|
| 1005 |
+
(anchor.insert_paragraph_before if anchor is not None else doc.add_paragraph)(
|
| 1006 |
+
f"[Attached figure '{caption}' could not be embedded — file unreadable.]")
|
| 1007 |
+
|
| 1008 |
+
|
| 1009 |
def generate_report_package(
|
| 1010 |
*,
|
| 1011 |
metadata: ReportMetadata,
|
|
|
|
| 1022 |
scenario_records: list[Mapping[str, Any]] | None = None,
|
| 1023 |
scenario_reporting_mode: str = "Base report with scenario comparison",
|
| 1024 |
preliminary_review_artifacts: Mapping[str, Any] | None = None,
|
| 1025 |
+
attached_figures: list[Mapping[str, Any]] | None = None,
|
| 1026 |
) -> dict[str, bytes | str]:
|
| 1027 |
"""Generate editable Word report and ZIP package entirely in memory."""
|
| 1028 |
criteria = criteria or ReportCriteria()
|
|
|
|
| 1390 |
"scenario_comparison": scenario_comparison if scenario_comparison is not None else pd.DataFrame(),
|
| 1391 |
},
|
| 1392 |
)
|
| 1393 |
+
docx_buffer = io.BytesIO()
|
| 1394 |
+
_embed_attached_figures(doc, attached_figures)
|
| 1395 |
+
doc.save(docx_buffer); docx_bytes = docx_buffer.getvalue()
|
| 1396 |
base = _safe_name(metadata.project_name)
|
| 1397 |
|
| 1398 |
zip_buffer = io.BytesIO()
|
|
|
|
| 1430 |
if frame is not None and not frame.empty:
|
| 1431 |
zf.writestr(f"tables/scenarios/{sid}_{key}.csv", frame.to_csv(index=False))
|
| 1432 |
zf.writestr("metadata/llm_report_context.json", json.dumps(llm_context, indent=2, default=str))
|
| 1433 |
+
if attached_figures:
|
| 1434 |
+
fig_manifest = []
|
| 1435 |
+
for n, fig in enumerate(attached_figures, start=1):
|
| 1436 |
+
fpath = Path(str(fig.get("path", "")))
|
| 1437 |
+
if fpath.exists():
|
| 1438 |
+
zf.writestr(f"figures/{fpath.name}", fpath.read_bytes())
|
| 1439 |
+
fig_manifest.append({"n": n, "figure_id": fig.get("figure_id"),
|
| 1440 |
+
"file": fpath.name, "caption": fig.get("caption"),
|
| 1441 |
+
"section": fig.get("section"), "source": fig.get("source"),
|
| 1442 |
+
"note": "Client-attached illustrative figure; not a server-verified output."})
|
| 1443 |
+
zf.writestr("metadata/attached_figures.json", json.dumps(fig_manifest, indent=2, default=str))
|
| 1444 |
zf.writestr("metadata/approved_narrative_sections.json", json.dumps(dict(narrative_sections or {}), indent=2, ensure_ascii=False))
|
| 1445 |
zf.writestr("metadata/project_metadata.json", json.dumps(asdict(metadata), indent=2))
|
| 1446 |
zf.writestr("metadata/report_criteria.json", json.dumps(asdict(criteria), indent=2, default=str))
|
requirements.txt
CHANGED
|
@@ -3,8 +3,8 @@ mcp>=1.10,<2
|
|
| 3 |
fastapi>=0.110
|
| 4 |
uvicorn>=0.29
|
| 5 |
httpx>=0.27
|
|
|
|
| 6 |
pandas>=2.0
|
| 7 |
numpy>=1.26
|
| 8 |
python-docx>=1.1
|
| 9 |
PyYAML>=6.0
|
| 10 |
-
requests>=2.31
|
|
|
|
| 3 |
fastapi>=0.110
|
| 4 |
uvicorn>=0.29
|
| 5 |
httpx>=0.27
|
| 6 |
+
requests>=2.31
|
| 7 |
pandas>=2.0
|
| 8 |
numpy>=1.26
|
| 9 |
python-docx>=1.1
|
| 10 |
PyYAML>=6.0
|
|
|
smoke_test.py
CHANGED
|
@@ -33,6 +33,14 @@ async def main():
|
|
| 33 |
print(f"MCP workflow: session {sid} | recon {recon.get('ok')}/{recon.get('links_checked')}")
|
| 34 |
if "Kincora" in INP:
|
| 35 |
assert recon.get("ok", 0) >= 26, "REGRESSION PIN FAILED"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
rep = json.loads((await s.call_tool("generate_report",
|
| 37 |
{"session_id": sid, "project_name": "Smoke Test"})).content[0].text)
|
| 38 |
dl = httpx.get(f"{BASE}{rep['files']['docx']}", timeout=60)
|
|
|
|
| 33 |
print(f"MCP workflow: session {sid} | recon {recon.get('ok')}/{recon.get('links_checked')}")
|
| 34 |
if "Kincora" in INP:
|
| 35 |
assert recon.get("ok", 0) >= 26, "REGRESSION PIN FAILED"
|
| 36 |
+
# 1x1 red PNG — validates the attach_figure -> embedded-report path
|
| 37 |
+
tiny_png = ("iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR4"
|
| 38 |
+
"2mP8z8BQDwAEhQGAhKmMIQAAAABJRU5ErkJggg==")
|
| 39 |
+
fig = json.loads((await s.call_tool("attach_figure",
|
| 40 |
+
{"session_id": sid, "image_base64": tiny_png,
|
| 41 |
+
"caption": "Smoke-test figure", "section": "results"})).content[0].text)
|
| 42 |
+
assert fig["figure_id"] == "FIG-01", fig
|
| 43 |
+
print("attach_figure: OK", fig["figure_id"])
|
| 44 |
rep = json.loads((await s.call_tool("generate_report",
|
| 45 |
{"session_id": sid, "project_name": "Smoke Test"})).content[0].text)
|
| 46 |
dl = httpx.get(f"{BASE}{rep['files']['docx']}", timeout=60)
|
tools.py
CHANGED
|
@@ -220,8 +220,13 @@ def get_timeseries(session_id: str, object_type: str, object_id: str, variable:
|
|
| 220 |
else:
|
| 221 |
points = [{"t": str(times[i]) if i < len(times) else i, "v": round(float(v), 6)}
|
| 222 |
for i, v in enumerate(series)]
|
|
|
|
|
|
|
|
|
|
| 223 |
return {"object_id": object_id, "variable": variable, "n_source_points": n,
|
| 224 |
-
"decimated": n > MAX_TS_POINTS, "peak": round(float(
|
|
|
|
|
|
|
| 225 |
"points": points}
|
| 226 |
|
| 227 |
|
|
@@ -349,6 +354,46 @@ def run_scenario(session_id: str, scenario_name: str,
|
|
| 349 |
"deterministic_analysis": narrative[:6000]}
|
| 350 |
|
| 351 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 352 |
def generate_report(session_id: str, project_name: str, client: str = "",
|
| 353 |
consultant: str = "", prepared_by: str = "",
|
| 354 |
outline_plan_no: str = "") -> dict:
|
|
@@ -372,7 +417,8 @@ def generate_report(session_id: str, project_name: str, client: str = "",
|
|
| 372 |
preliminary_review_artifacts={
|
| 373 |
"findings": findings, "status": "Preliminary",
|
| 374 |
"manifest": {"rpt_reconciliation": session.data.get("recon_summary", {})},
|
| 375 |
-
} if findings else None
|
|
|
|
| 376 |
outputs = session.workdir / "outputs"
|
| 377 |
outputs.mkdir(exist_ok=True)
|
| 378 |
files = {}
|
|
@@ -399,6 +445,6 @@ TOOL_REGISTRY: dict[str, Callable[..., dict]] = {
|
|
| 399 |
get_node_results, get_link_results, get_subcatchment_results,
|
| 400 |
get_timeseries, query_results, get_table_catalog,
|
| 401 |
calgary_screening, preliminary_design_review, get_reconciliation,
|
| 402 |
-
run_scenario, generate_report,
|
| 403 |
]
|
| 404 |
}
|
|
|
|
| 220 |
else:
|
| 221 |
points = [{"t": str(times[i]) if i < len(times) else i, "v": round(float(v), 6)}
|
| 222 |
for i, v in enumerate(series)]
|
| 223 |
+
peak_val = max(series, key=abs, default=0.0)
|
| 224 |
+
peak_idx = series.index(peak_val) if series else 0
|
| 225 |
+
time_of_peak = str(times[peak_idx]) if peak_idx < len(times) else None
|
| 226 |
return {"object_id": object_id, "variable": variable, "n_source_points": n,
|
| 227 |
+
"decimated": n > MAX_TS_POINTS, "peak": round(float(peak_val), 6),
|
| 228 |
+
"time_of_peak": time_of_peak,
|
| 229 |
+
"note": "time_of_peak is from the full-resolution series; do not infer it from decimated point labels.",
|
| 230 |
"points": points}
|
| 231 |
|
| 232 |
|
|
|
|
| 354 |
"deterministic_analysis": narrative[:6000]}
|
| 355 |
|
| 356 |
|
| 357 |
+
def attach_figure(session_id: str, image_base64: str, caption: str,
|
| 358 |
+
section: str = "results", figure_name: str = "") -> dict:
|
| 359 |
+
"""Attach a client-generated figure (PNG/JPEG, base64) to the session so
|
| 360 |
+
generate_report embeds it in the AUDITED report instead of the client
|
| 361 |
+
rebuilding the document itself.
|
| 362 |
+
|
| 363 |
+
section: keyword matched against report Heading-1 titles (e.g. "results",
|
| 364 |
+
"methodology", "site"); the figure is placed at the end of that section.
|
| 365 |
+
Figures are labelled as client-attached illustrative material — they are
|
| 366 |
+
not server-verified outputs. Limits: PNG or JPEG, 5 MB decoded.
|
| 367 |
+
"""
|
| 368 |
+
session = STORE.get(session_id)
|
| 369 |
+
try:
|
| 370 |
+
blob = base64.b64decode(image_base64, validate=True)
|
| 371 |
+
except (binascii.Error, ValueError) as exc:
|
| 372 |
+
raise ValueError(f"image_base64 is not valid base64: {exc}")
|
| 373 |
+
if len(blob) > 5 * 1024 * 1024:
|
| 374 |
+
raise ValueError("Figure exceeds the 5 MB limit.")
|
| 375 |
+
if blob[:8] == b"\x89PNG\r\n\x1a\n":
|
| 376 |
+
ext = "png"
|
| 377 |
+
elif blob[:3] == b"\xff\xd8\xff":
|
| 378 |
+
ext = "jpg"
|
| 379 |
+
else:
|
| 380 |
+
raise ValueError("Only PNG or JPEG figures are accepted (magic-byte check failed).")
|
| 381 |
+
figures = session.data.setdefault("figures", [])
|
| 382 |
+
figure_id = f"FIG-{len(figures) + 1:02d}"
|
| 383 |
+
safe = "".join(c if c.isalnum() or c in "-_" else "_" for c in (figure_name or figure_id))
|
| 384 |
+
fig_dir = session.workdir / "figures"
|
| 385 |
+
fig_dir.mkdir(exist_ok=True)
|
| 386 |
+
path = fig_dir / f"{safe}.{ext}"
|
| 387 |
+
path.write_bytes(blob)
|
| 388 |
+
figures.append({"figure_id": figure_id, "path": str(path), "caption": caption.strip(),
|
| 389 |
+
"section": section.strip().lower() or "results", "source": "client-attached"})
|
| 390 |
+
return {"figure_id": figure_id, "stored": path.name, "size_bytes": len(blob),
|
| 391 |
+
"section": section, "attached_figures": [
|
| 392 |
+
{"figure_id": f["figure_id"], "caption": f["caption"], "section": f["section"]}
|
| 393 |
+
for f in figures],
|
| 394 |
+
"next_step": "Call generate_report; the figure will be embedded with a labelled caption."}
|
| 395 |
+
|
| 396 |
+
|
| 397 |
def generate_report(session_id: str, project_name: str, client: str = "",
|
| 398 |
consultant: str = "", prepared_by: str = "",
|
| 399 |
outline_plan_no: str = "") -> dict:
|
|
|
|
| 417 |
preliminary_review_artifacts={
|
| 418 |
"findings": findings, "status": "Preliminary",
|
| 419 |
"manifest": {"rpt_reconciliation": session.data.get("recon_summary", {})},
|
| 420 |
+
} if findings else None,
|
| 421 |
+
attached_figures=session.data.get("figures") or None)
|
| 422 |
outputs = session.workdir / "outputs"
|
| 423 |
outputs.mkdir(exist_ok=True)
|
| 424 |
files = {}
|
|
|
|
| 445 |
get_node_results, get_link_results, get_subcatchment_results,
|
| 446 |
get_timeseries, query_results, get_table_catalog,
|
| 447 |
calgary_screening, preliminary_design_review, get_reconciliation,
|
| 448 |
+
run_scenario, attach_figure, generate_report,
|
| 449 |
]
|
| 450 |
}
|