panda
bd5bfbac2d
Root cause: LLM receiving full 34k-char JRXML would regenerate from scratch
instead of modifying coordinates in-place, shrinking output to ~3k chars.
Solution (programmatic node control, not prompt engineering):
- New agent/jrxml_windower.py: decompose JRXML into header (never sent to
LLM) + individual bands. Split bands >4000 chars at element boundaries.
Reassemble with element count validation (>10% change = rollback).
- Rewrite refine_layout: per-band windowed LLM processing (~2-4k chars
each). LLM cannot "reimagine" the entire report.
- Rewrite map_fields: 100% programmatic regex $F{field_N} -> real name
replacement. Zero LLM calls, zero content loss.
- _sanitize_field_name: non-ASCII chars escaped to _uXXXX_ format for
valid JRXML identifiers.
- Tests: 48 new unit tests (windower 28 + map_fields 20). All passing.
Full suite 385 tests, zero regressions.
13 lines
346 B
JSON
13 lines
346 B
JSON
{
|
|
"kb_id": "49b972ec9e424f04aec34899c978f087",
|
|
"user_id": "2db10c2ebbf6434aab28035026e196c3",
|
|
"name": "smoke_kb",
|
|
"description": "",
|
|
"created_at": "2026-05-23T12:21:32.409028+00:00",
|
|
"updated_at": "2026-05-23T12:21:32.409028+00:00",
|
|
"fields": [],
|
|
"templates": [],
|
|
"file_count": 0,
|
|
"chunk_count": 0,
|
|
"parse_status": "empty"
|
|
} |