#!/usr/bin/env python3
"""Render the local EP1 report from authoritative snapshots and supplementary checks.
Never changes snapshots, runs, scores or credentials. Output works over file://.
"""
import hashlib, json, pathlib, runpy
ROOT=pathlib.Path(__file__).resolve().parents[1]
TEMPLATES=ROOT/'site/templates'
def main():
    files=[ROOT/'data/experiments/exp-20260915-ep1-f005-docker.json',ROOT/'data/experiments/exp-20260918-ep1-f005-glm-probe1200.json']
    snapshots=[json.loads(p.read_text()) for p in files]
    evidence=json.loads((ROOT/'site/evidence/ep1-reference/comparison.json').read_text())
    cases={c['id']:c for c in evidence['runs']}
    rows=[]
    for snap in snapshots:
        for raw in snap['runs']:
            if raw['task']!='ep1-f005':continue
            u=raw.get('wire_backfill') or raw['official_usage']
            net=u.get('net_input',u.get('input_tokens')); cache=u.get('cache_read',u.get('cache_read_tokens'))
            row={
                'id':raw['run_id'],'model':raw['model'],'attempt':raw['attempt'],'timestamp':raw['ts'],
                'helper':cases[raw['run_id']]['helper'],'sourceSha256':cases[raw['run_id']]['sha256'],
                'budget':raw['params']['timeout_s'],'seconds':raw['wall_s'],'score':raw['score'],'exit':raw['agent_exit'],
                'calls':u['api_calls'],'net':net,'cache':cache,'output':u.get('output',u.get('output_tokens')),
                'usageSource':'wire' if raw.get('wire_backfill') else 'official',
                'testModified':raw['score_diag'].get('test_modified_by_model'),
                'checks':cases[raw['run_id']]['checks'],'mechanism':cases[raw['run_id']]['mechanism'],
                'diff':(ROOT/'site/evidence/ep1-reference'/f"{raw['run_id']}.diff").read_text(),
            }
            assert all(row[k] is not None for k in ['net','cache','calls','output'])
            rows.append(row)
    rows.sort(key=lambda r:(r['budget'],r['model'],r['attempt']))
    main_runs=[r for r in rows if r['budget']==420]
    passed=[r for r in rows if r['score']==1]
    main_models=set(r['model'] for r in main_runs)
    summary={
        'mainRuns':len(main_runs),'mainModels':len(main_models),'mainPassed':sum(r['score']==1 for r in main_runs),
        'mainTimeouts':sum(r['exit']==124 for r in main_runs),
        'mainModelsBothPassed':sum(all(r['score']==1 for r in main_runs if r['model']==m) for m in main_models),
        'allTaskRuns':len(rows),'passedRuns':len(passed),
        'passedCopyPass':sum(r['checks']['api_key_copy']['pass'] for r in passed),
        'checksRequests':sum(len(c['checks']) for c in cases.values()),
    }
    assert len(rows)==20 and len(main_runs)==18 and len(main_models)==9
    assert summary['mainPassed']==14 and summary['passedCopyPass']==0
    answer_review=json.loads((ROOT/'site/evidence/ep1-reference/answer-review.json').read_text())
    assert {r['id'] for r in answer_review['runs']}=={r['id'] for r in rows}
    assert sum(r['status']=='final' for r in answer_review['runs'])==16
    for item in answer_review['runs']+answer_review['highlights']:
        source=ROOT/item['source']
        assert hashlib.sha256(source.read_bytes()).hexdigest()==item['sourceSha256'], str(source)
        body=json.loads(source.read_text())['body']
        visible=''
        for line in body.splitlines():
            if not line.startswith('data: '):continue
            try: chunk=json.loads(line[6:])
            except json.JSONDecodeError:continue
            for choice in chunk.get('choices',[]):
                visible+=choice.get('delta',choice.get('message',{})).get('content') or ''
        assert item.get('quote',item.get('text')) in visible, str(source)
    answer_corpus=json.loads((ROOT/'site/evidence/ep1-reference/answer-corpus.json').read_text())
    answer_rubric=json.loads((ROOT/'site/evidence/ep1-reference/answer-rubric.json').read_text())
    run_ids={r['id'] for r in rows}
    assert {r['id'] for r in answer_corpus['runs']}==run_ids
    assert {r['id'] for r in answer_rubric['runs']}==run_ids
    parse_reply=runpy.run_path(str(ROOT/'scripts/extract-ep1-answers.py'))['visible_reply']
    for run in answer_corpus['runs']:
        for message in run['messages']:
            source=ROOT/message['source']
            assert hashlib.sha256(source.read_bytes()).hexdigest()==message['sourceSha256']
            text,tools=parse_reply(source)
            assert text==message['text'] and tools==message['hasToolCalls']
    for item in answer_rubric['runs']:
        messages=next(r['messages'] for r in answer_corpus['runs'] if r['id']==item['id'])
        assert item['reviewedMessages']==len(messages)
        assert all(e in messages for e in item['evidence'])
    data={
        'answerReview':answer_review,'answerCorpus':answer_corpus,'answerRubric':answer_rubric,
        'runs':rows,'summary':summary,'fixture':snapshots[0]['fixture'],
        'reference':cases['upstream-reference'],'baseline':cases['baseline'],
        'checkDate':evidence['generated_at'],'sdk':evidence['anthropic_version'],
        'findings':json.loads((TEMPLATES/'ep1-findings.json').read_text()),
        'sourceHashes':{str(p.relative_to(ROOT)):hashlib.sha256(p.read_bytes()).hexdigest() for p in files},
    }
    payload=json.dumps(data,ensure_ascii=False,separators=(',',':')).replace('<','\\u003c').replace('>','\\u003e').replace('&','\\u0026')
    for template,filename in [('ep1-report.html','ep1-douban.html'),('ep1-evidence.html','ep1-evidence.html')]:
        out=(TEMPLATES/template).read_text().replace('/* REPORT_TOKENS */',(TEMPLATES/'ep1-tokens.css').read_text()).replace('/* REPORT_DATA */',payload)
        (ROOT/'site/drafts'/filename).write_text(out)
    print(json.dumps(summary,ensure_ascii=False))
if __name__=='__main__':main()