1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714"""Document processing endpoints."""
import asyncio
import json
import logging
from collections.abc import AsyncGenerator
from datetime import UTC, datetime
from typing import Any, Literal
from fastapi import APIRouter, BackgroundTasks, Depends, File, Form, HTTPException, Query, UploadFile, status
from fastapi.responses import StreamingResponse
from pydantic import BaseModel, Field
from ..config import settings
from ..dependencies import (
get_job_service,
get_queue_service,
get_redis_client,
get_s3_url_service,
get_storage_service,
)
from ..services import JobService, QueueService, S3URLService, StorageService
from ..services.document_processing_service import DocumentProcessingService
from ..services.metrics_service import jobs_submitted_total
from .schemas import (
AgenticCompletedResponse,
AgenticProcessingResponse,
AwaitingPIIApprovalResponse,
CompletedResponse,
DeniedResponse,
DocumentStatusResponse,
FailedResponse,
FigureAsset,
LedgerEntryResponse,
LedgerPageGroup,
LedgerResponse,
LLMCallInfo,
LLMCostInfo,
PIIFinding,
PIIScanningResponse,
ProcessingResponse,
)
logger = logging.getLogger(__name__)
router = APIRouter(prefix="/api/v1/documents", tags=["Documents"])
def _build_llm_cost(job: dict[str, Any]) -> LLMCostInfo | None:
"""Build LLM cost info from job data.
Costs accumulate across all processing phases (structure analysis + transcription).
"""
llm_cost_cents = job.get("llm_cost_cents")
if llm_cost_cents is None:
return None
try:
total_cents = float(llm_cost_cents)
except (TypeError, ValueError):
return None
# Get aggregate token counts (accumulated across all phases)
input_tokens = int(job.get("llm_input_tokens", 0) or 0)
output_tokens = int(job.get("llm_output_tokens", 0) or 0)
total_tokens = int(job.get("llm_total_tokens", 0) or 0)
# Parse individual LLM calls from job data
llm_calls_raw = job.get("llm_calls")
calls: list[LLMCallInfo] = []
if llm_calls_raw:
# llm_calls is stored as JSON string in Redis
if isinstance(llm_calls_raw, str):
try:
llm_calls_raw = json.loads(llm_calls_raw)
except json.JSONDecodeError:
llm_calls_raw = []
if isinstance(llm_calls_raw, list):
for call in llm_calls_raw:
if isinstance(call, dict):
calls.append(
LLMCallInfo(
agent=call.get("agent", ""),
purpose=call.get("purpose", ""),
page=call.get("page"),
input_tokens=call.get("input_tokens", 0),
output_tokens=call.get("output_tokens", 0),
cost_cents=call.get("cost_cents", 0.0),
timestamp=call.get("timestamp", ""),
duration_ms=call.get("duration_ms"),
)
)
return LLMCostInfo(
input_tokens=input_tokens,
output_tokens=output_tokens,
total_tokens=total_tokens,
estimated_cost_cents=total_cents,
estimated_cost_dollars=total_cents / 100.0,
calls=calls,
)
class JobSubmissionResponse(BaseModel):
"""Response for document submission."""
job_id: str
status: str
estimated_completion_minutes: int
created_at: str
stream_url: str | None = None
@router.post("/submit", response_model=JobSubmissionResponse, status_code=status.HTTP_201_CREATED)
async def submit_document(
background_tasks: BackgroundTasks,
file: UploadFile = File(...),
skip_pii_scan: bool = Form(default=False, description="Skip PII scanning and use agentic pipeline directly"),
skip_reason: str | None = Form(default=None, description="Optional reason for skipping PII scan (for audit trail)"),
review_mode: Literal["auto", "human"] = Form(
default="auto",
description="Review mode: 'auto' (immediate completion) or 'human' (ledger available for PR-like review)",
),
generate_debug_bundle: bool = Form(
default=False, description="Generate debug bundle with all agent prompts and responses"
),
ocr_languages: str | None = Form(
default=None,
description="Comma-separated Tesseract OCR language codes for scanned documents (e.g. 'eng,deu'). Defaults to 'eng'.",
),
storage: StorageService = Depends(get_storage_service),
queue: QueueService = Depends(get_queue_service),
job_service: JobService = Depends(get_job_service),
redis_client: Any = Depends(get_redis_client),
s3_url_service: S3URLService = Depends(get_s3_url_service),
) -> JobSubmissionResponse:
"""Submit a PDF document for processing.
Args:
file: PDF file to process
skip_pii_scan: If True, bypass PII scanning and use agentic pipeline directly
skip_reason: Optional justification for skipping PII scan (recorded in audit trail)
review_mode: 'auto' (immediate completion) or 'human' (ledger available for review)
generate_debug_bundle: If True, save all agent prompts/responses for debugging
"""
# Parse and validate OCR languages
ocr_lang_list = [lang.strip() for lang in ocr_languages.split(",")] if ocr_languages else None
if ocr_lang_list:
from ..utils.ocr_languages import validate_ocr_languages
invalid = validate_ocr_languages(ocr_lang_list)
if invalid:
raise HTTPException(
status_code=status.HTTP_422_UNPROCESSABLE_ENTITY,
detail=f"Invalid OCR language code(s): {', '.join(invalid)}. Use Tesseract language codes (e.g. 'eng', 'deu', 'fra').",
)
job_id, s3_key = await storage.store_document(file)
if skip_pii_scan:
# Use agentic pipeline directly (bypass PII scanning)
await job_service.create_job(
job_id,
s3_key,
status="processing",
original_filename=file.filename,
pii_skipped=True,
pii_skip_reason=skip_reason or "User requested PII scan skip",
debug_bundle_requested=generate_debug_bundle,
review_mode=review_mode,
ocr_languages=ocr_lang_list,
)
# Pre-warm docling (best-effort, production only)
if settings.environment == "production":
try:
import boto3
cw = boto3.client("cloudwatch", region_name=settings.aws_region)
cw.put_metric_data(
Namespace="EqualifyPDF",
MetricData=[{"MetricName": "JobsAwaitingProcessing", "Value": 1, "Unit": "Count"}],
)
except Exception:
pass
# Record job submission metric
jobs_submitted_total.labels(source="api").inc()
# Use DocumentProcessingService for agentic pipeline
processing_service = DocumentProcessingService(
redis_client=redis_client,
storage_service=storage,
s3_url_service=s3_url_service,
)
# Run processing in background
background_tasks.add_task(
processing_service.process_document,
job_id=job_id,
s3_key=s3_key,
filename=file.filename or "document.pdf",
review_mode=review_mode,
ocr_languages=ocr_lang_list,
)
return JobSubmissionResponse(
job_id=job_id,
status="processing",
estimated_completion_minutes=settings.estimated_processing_minutes,
created_at=datetime.now(UTC).isoformat().replace("+00:00", "Z"),
stream_url=f"/api/v1/documents/{job_id}/stream",
)
else:
# Standard flow: PII scanning first
await job_service.create_job(
job_id,
s3_key,
status="pii_scanning",
original_filename=file.filename,
debug_bundle_requested=generate_debug_bundle,
review_mode=review_mode,
ocr_languages=ocr_lang_list,
)
await queue.queue_pii_job(job_id, s3_key)
# Pre-warm docling (best-effort, production only)
if settings.environment == "production":
try:
import boto3
cw = boto3.client("cloudwatch", region_name=settings.aws_region)
cw.put_metric_data(
Namespace="EqualifyPDF",
MetricData=[{"MetricName": "JobsAwaitingProcessing", "Value": 1, "Unit": "Count"}],
)
except Exception:
pass
# Record job submission metric
jobs_submitted_total.labels(source="api").inc()
return JobSubmissionResponse(
job_id=job_id,
status="pii_scanning",
estimated_completion_minutes=settings.estimated_processing_minutes,
created_at=datetime.now(UTC).isoformat().replace("+00:00", "Z"),
)
@router.get("/{job_id}", response_model=DocumentStatusResponse)
async def get_job(
job_id: str,
job_service: JobService = Depends(get_job_service),
url_service: S3URLService = Depends(get_s3_url_service),
) -> DocumentStatusResponse:
"""
Get current status of a processing job.
Returns a clean, status-specific response with only relevant fields.
"""
job = await job_service.get_job(job_id)
if not job:
raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=f"Job {job_id} not found")
base = {
"job_id": job["job_id"],
"status": job["status"],
"filename": job.get("original_filename"),
"created_at": job["created_at"],
"updated_at": job["updated_at"],
"debug_bundle_requested": job.get("debug_bundle_requested") == "true",
}
match job["status"]:
case "pii_scanning":
return PIIScanningResponse(
**base,
estimated_completion_minutes=settings.estimated_processing_minutes,
)
case "processing":
# Check if this is an agentic pipeline job (has review_mode set)
review_mode = job.get("review_mode")
processing_phase = job.get("processing_phase")
if review_mode:
# Agentic pipeline response
return AgenticProcessingResponse(
**base,
review_mode=review_mode,
processing_phase=processing_phase or "initializing",
jobs_total=int(job.get("jobs_total", 0)),
jobs_complete=int(job.get("jobs_complete", 0)),
stream_url=f"/api/v1/documents/{job_id}/stream",
pii_skipped=job.get("pii_skipped") == "true" if job.get("pii_skipped") else None,
)
else:
# Legacy pipeline response
return ProcessingResponse(
**base,
estimated_completion_minutes=settings.estimated_processing_minutes,
pii_skipped=job.get("pii_skipped") == "true" if job.get("pii_skipped") else None,
)
case "awaiting_approval":
pii_findings = [PIIFinding(**f) for f in (job.get("pii_findings") or [])]
token = job.get("approval_token", "")
return AwaitingPIIApprovalResponse(
**base,
pii_findings=pii_findings,
approval_token=token,
approval_expires_at=job.get("approval_expires_at", ""),
approval_url=f"/api/v1/approval/{token}/decision",
)
case "completed":
# Check if this is an agentic pipeline job (has review_mode set)
review_mode = job.get("review_mode")
if review_mode:
# Agentic pipeline completed response
# Get result_url (markdown) - different field name for agentic pipeline
markdown_key = job.get("result_url") or job.get("markdown_url")
if not markdown_key:
raise HTTPException(
status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
detail="Job completed but result_url/markdown_url not set. This indicates a bug.",
)
# Ledger is always available for completed agentic pipeline jobs
ledger_url = f"/api/v1/documents/{job_id}/ledger"
# Build figure assets from stored_figures
figure_assets: list[FigureAsset] = []
stored_figures_raw = job.get("stored_figures")
if stored_figures_raw:
# Parse if stored as JSON string
if isinstance(stored_figures_raw, str):
try:
stored_figures_data = json.loads(stored_figures_raw)
except json.JSONDecodeError:
stored_figures_data = []
elif isinstance(stored_figures_raw, bytes):
try:
stored_figures_data = json.loads(stored_figures_raw.decode("utf-8"))
except (json.JSONDecodeError, UnicodeDecodeError):
stored_figures_data = []
else:
stored_figures_data = stored_figures_raw
for fig in stored_figures_data:
s3_key = fig.get("s3_key", "")
if s3_key:
figure_url = await url_service.generate_url(
s3_key, bucket=url_service.results_bucket
)
figure_assets.append(
FigureAsset(
figure_id=fig.get("figure_id", ""),
url=figure_url,
page=int(fig.get("page_num", 1)),
alt_text=fig.get("alt_text", ""),
caption=fig.get("caption", ""),
)
)
# Bundle URL is only available if there are figures
bundle_url = (
f"/api/v1/documents/{job_id}/bundle"
if figure_assets
else None
)
return AgenticCompletedResponse(
**base,
review_mode=review_mode,
markdown_url=await url_service.generate_url(markdown_key, bucket=url_service.results_bucket),
confidence_score=float(job.get("confidence_score", 0.0)),
llm_cost=_build_llm_cost(job)
or LLMCostInfo(
input_tokens=0,
output_tokens=0,
total_tokens=0,
estimated_cost_cents=0,
estimated_cost_dollars=0,
),
ledger_url=ledger_url,
total_pages=int(job.get("total_pages", 0)),
total_edits=int(job.get("total_edits", 0)),
figures=figure_assets,
bundle_url=bundle_url,
)
else:
# Legacy pipeline completed response
# PRD-027: markdown_url must be saved in job record by apply_reviews
markdown_key = job.get("markdown_url")
if not markdown_key:
raise HTTPException(
status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
detail="Job completed but markdown_url not set. This indicates a bug.",
)
return CompletedResponse(
**base,
markdown_url=await url_service.generate_url(markdown_key, bucket=url_service.results_bucket),
confidence_score=float(job.get("confidence_score", 0.0)),
llm_cost=_build_llm_cost(job)
or LLMCostInfo(
input_tokens=0,
output_tokens=0,
total_tokens=0,
estimated_cost_cents=0,
estimated_cost_dollars=0,
),
)
case "failed":
return FailedResponse(
**base,
error=job.get("error", "Unknown error"),
)
case "denied":
return DeniedResponse(
**base,
reason=job.get("denial_reason", "PII not approved"),
)
case _:
raise HTTPException(
status_code=status.HTTP_500_INTERNAL_SERVER_ERROR,
detail=f"Unknown job status: {job['status']}",
)
class StreamTokenResponse(BaseModel):
"""Response for stream token generation."""
token: str = Field(..., description="Single-use token for SSE authentication")
expires_in_seconds: int = Field(300, description="Token validity period in seconds")
stream_url: str = Field(..., description="Full URL path with token for EventSource")
@router.post("/{job_id}/stream/token", response_model=StreamTokenResponse)
async def create_stream_token(
job_id: str,
job_service: JobService = Depends(get_job_service),
) -> StreamTokenResponse:
"""Generate single-use token for SSE stream authentication.
Browser EventSource API cannot send custom headers. This endpoint
generates a short-lived, single-use token that can be passed as
a query parameter to the stream endpoint.
The token:
- Expires in 5 minutes
- Is consumed on first use (single-use)
- Is scoped to the specific job_id
Args:
job_id: Job identifier
Returns:
StreamTokenResponse with token and stream URL
Raises:
HTTPException 404: Job not found
"""
# Verify job exists
job = await job_service.get_job(job_id)
if not job:
raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=f"Job {job_id} not found")
# Generate single-use token
token = await job_service.create_stream_token(job_id)
return StreamTokenResponse(
token=token,
expires_in_seconds=300,
stream_url=f"/api/v1/documents/{job_id}/stream?token={token}",
)
@router.get("/{job_id}/stream")
async def stream_events(
job_id: str,
token: str | None = Query(None, description="Single-use stream token from /stream/token endpoint"),
job_service: JobService = Depends(get_job_service),
) -> StreamingResponse:
"""Stream processing events via Server-Sent Events (SSE).
Authentication: Either X-API-Key header OR valid stream token.
Stream tokens are generated via POST /{job_id}/stream/token.
Connect to this endpoint to watch processing in real-time.
Events include: docling progress, planning progress, job creation, edits, etc.
Args:
job_id: Job identifier
token: Optional single-use stream token (for browser EventSource)
Returns:
SSE stream with processing events
Raises:
HTTPException 401: Invalid or missing authentication
HTTPException 404: Job not found
"""
# If token provided, validate and consume it (single-use)
if token:
validated_job_id = await job_service.validate_and_consume_stream_token(token)
if not validated_job_id:
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
detail="Invalid, expired, or already-used stream token",
)
# Verify token is for this job (job-scoped security)
if validated_job_id != job_id:
raise HTTPException(
status_code=status.HTTP_401_UNAUTHORIZED,
detail="Stream token is not valid for this job",
)
# Note: If no token, middleware already validated API key
from ..agents.events import get_event_bus
# Verify job exists
job = await job_service.get_job(job_id)
if not job:
raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=f"Job {job_id} not found")
async def event_generator() -> AsyncGenerator[str, None]:
"""Generate SSE events."""
try:
# Get event bus from registry
event_bus = get_event_bus(job_id)
# If job is already complete or failed, send final event and close
if job["status"] in ("completed", "failed"):
if event_bus:
for event in event_bus.events:
yield f"event: {event.event_type}\ndata: {event.model_dump_json()}\n\n"
yield "event: done\ndata: {}\n\n"
return
# Wait for event bus to be available (up to 30 seconds)
max_wait = 30.0
waited = 0.0
while event_bus is None and waited < max_wait:
await asyncio.sleep(0.5)
waited += 0.5
event_bus = get_event_bus(job_id)
if event_bus is None:
yield 'event: error\ndata: {"message": "Event bus not available"}\n\n'
yield "event: done\ndata: {}\n\n"
return
# Subscribe to events
queue = event_bus.subscribe()
try:
# First, send any events that already happened
for event in event_bus.events:
yield f"event: {event.event_type}\ndata: {event.model_dump_json()}\n\n"
# Then stream new events
while True:
try:
event = await asyncio.wait_for(queue.get(), timeout=30.0)
yield f"event: {event.event_type}\ndata: {event.model_dump_json()}\n\n"
# Check if processing is complete
if event.event_type in ("processing:complete", "processing:error"):
break
except TimeoutError:
# Send keepalive
yield ": keepalive\n\n"
# Check if job is done by refreshing status
refreshed_job = await job_service.get_job(job_id)
if refreshed_job and refreshed_job["status"] in ("completed", "failed"):
break
finally:
event_bus.unsubscribe(queue)
yield "event: done\ndata: {}\n\n"
except Exception as e:
logger.error(f"SSE stream error for job {job_id}: {e}")
yield f'event: error\ndata: {{"message": "Stream error: {str(e)}"}}\n\n'
yield "event: done\ndata: {}\n\n"
return StreamingResponse(
event_generator(),
media_type="text/event-stream",
headers={
"Cache-Control": "no-cache",
"Connection": "keep-alive",
"X-Accel-Buffering": "no",
},
)
@router.get("/{job_id}/ledger", response_model=LedgerResponse)
async def get_ledger(
job_id: str,
job_service: JobService = Depends(get_job_service),
storage: StorageService = Depends(get_storage_service),
url_service: S3URLService = Depends(get_s3_url_service),
) -> LedgerResponse:
"""Get change ledger for PR-like review.
Returns the complete change ledger with all edits made by the pipeline,
grouped by page for easy review.
Args:
job_id: Job identifier
Returns:
LedgerResponse with all changes grouped by page
Raises:
HTTPException 404: Job not found or ledger not available
HTTPException 400: Job not yet complete
"""
# Verify job exists
job = await job_service.get_job(job_id)
if not job:
raise HTTPException(status_code=status.HTTP_404_NOT_FOUND, detail=f"Job {job_id} not found")
# Require job to be completed
if job["status"] != "completed":
raise HTTPException(
status_code=status.HTTP_400_BAD_REQUEST,
detail=f"Job not yet complete (status: {job['status']})",
)
# Get ledger from S3 via DocumentProcessingService
processing_service = DocumentProcessingService(
redis_client=None, # Not needed for get_ledger
storage_service=storage,
s3_url_service=url_service,
)
ledger_data = await processing_service.get_ledger(job_id)
if not ledger_data:
raise HTTPException(
status_code=status.HTTP_404_NOT_FOUND,
detail="Ledger not found for this job",
)
# Build grouped ledger response
entries_by_page: dict[int, list[LedgerEntryResponse]] = {}
entries_needing_review_count = 0
for entry in ledger_data.get("entries", []):
page = entry.get("page", 1)
if page not in entries_by_page:
entries_by_page[page] = []
needs_review = entry.get("needs_review", False)
if needs_review:
entries_needing_review_count += 1
entries_by_page[page].append(
LedgerEntryResponse(
entry_id=entry.get("entry_id", ""),
page=page,
action=entry.get("action", ""),
target=entry.get("target", ""),
before=entry.get("before", ""),
after=entry.get("after", ""),
reasoning=entry.get("reasoning", ""),
confidence=float(entry.get("confidence", 0.0)),
timestamp=entry.get("timestamp", ""),
needs_review=needs_review,
)
)
# Build page groups
pages = []
for page_num in sorted(entries_by_page.keys()):
page_entries = entries_by_page[page_num]
pages.append(
LedgerPageGroup(
page=page_num,
entries=page_entries,
edit_count=len(page_entries),
)
)
# Generate markdown URL
markdown_s3_key = job.get("result_url", f"results/{job_id}/result.md")
final_markdown_url = await url_service.generate_url(
markdown_s3_key,
bucket=url_service.results_bucket,
)
return LedgerResponse(
job_id=job_id,
document_title=job.get("original_filename", ""),
total_pages=int(job.get("total_pages", 0)),
pages_with_changes=len(pages),
total_edits=ledger_data.get("total_edits", 0),
entries_needing_review=entries_needing_review_count,
pages=pages,
processing_duration_ms=0,
final_markdown_url=final_markdown_url,
)