• Country: cl (Chile), es (Spain), pe (Peru), eu (European Union)
• Dates: Time range (last week, this month, specific date)
• Committee/Body: If a specific one is mentioned
• Analysis type: Summary, detailed, comparative, trends

curl -s "https://parlamento.ai/api/external/research/official-journal?country=COUNTRY&search=TERM&dateFrom=DATE&limit=20" \
  -H "Authorization: Bearer $PARLAMENTO_API_KEY"

Note: Peru (pe) does not have Official Journal integrated yet.

/official-journal

{
  "success": true,
  "data": {
    "country": "cl",
    "count": 5,
    "publications": [
      {
        "id": "oj-cl:2756348",
        "title": "Approves bike lane project Short Term Network Construction...",
        "documentType": "Resolution",
        "ministry": "Ministry of Transport and Telecommunications",
        "pdfUrl": "https://www.diariooficial.interior.gob.cl/publicaciones/...",
        "publishDate": "2026-01-20",
        "hasExtractedText": true,
        "textLength": 8252,
        "textPreview": "OFFICIAL JOURNAL OF THE REPUBLIC OF CHILE..."
      }
    ]
  }
}

• User asks about "energy decrees"
• User asks about "laws published this week"
• User asks about "Ministry of Health resolutions"
• User mentions "Official Journal", "BOE", "EUR-Lex"
• User wants to know what regulations have been published

• Add a "Related Regulations" or "Official Journal Publications" section
• Include title, document type, ministry, and date
• Use

  textPreview

  to give context about the content
• Include link to PDF:

  <a href="PDF_URL">View official document</a>

1. Count sessions found:

   SESSION_INVENTORY = []
   For each session in /transcripts response:
     - Add: { id, title, date, group, subgroup }
   TOTAL_SESSIONS = len(SESSION_INVENTORY)

2. Count Official Journal publications (if applicable):

   OJ_INVENTORY = []
   For each publication in /official-journal response:
     - Add: { id, title, date, documentType }
   TOTAL_OJ = len(OJ_INVENTORY)

3. Save these numbers - you'll use them in Phase 5.5 to validate:

   • TOTAL_SESSIONS

     : Exact number of sessions

   • TOTAL_OJ

     : Exact number of OJ publications

   • SESSION_INVENTORY

     : Complete list with IDs

   • OJ_INVENTORY

     : Complete list with IDs

1. Main topics identification
   • What issues were discussed?
   • What were the debate points?
2. Relevant quotes extraction
   • Important statements from parliamentarians
   • Positions of different political groups
3. Trend analysis
   • Are there patterns across sessions?
   • How did the discussion evolve?
4. Conclusions
   • Executive summary
   • Key points for the reader

1. Determine the country from the data you collected (cl, es, eu, pe)
2. Select the appropriate template (MANDATORY match):
   • Chile (cl): Read templates/html-template-cl.md
   • España (es): Read templates/html-template-es.md
   • European Union (eu): Read templates/html-template-eu.md
   • Peru (pe): Read templates/html-template-pe.md
   • Other countries: Read templates/html-template.md (legacy template, ONLY for countries without a specific template)
3. Read the selected template file to get the complete HTML structure
4. Replace ALL placeholders in the template:

   • {{DATE}}

     : Report creation date (e.g., "Febrero 2026")

   • {{TITLE}}

     : Descriptive report title (e.g., "Regulación de Inteligencia Artificial en la UE")

   • {{SUBTITLE}}

     : Subtitle or description of the report (e.g., "Análisis de sesiones plenarias y comités")

   • {{YEAR}}

     : Current year (2026)

   • {{CONTENT}}

     : Your complete analysis in HTML (use h2, h3, p, ul, blockquote, table)
   Note: Images (country flags, logos) are loaded automatically from URLs in the template. No base64 conversion needed.

If analyzing Chile's Health Committee:
  → countryCode = "cl"
  → Use templates/html-template-cl.md
  → Cover shows Chilean parliament flag image

If analyzing European Parliament:
  → countryCode = "eu"
  → Use templates/html-template-eu.md
  → Cover shows EU parliament flag image

1. Extract metrics from the HTML you wrote:
   • How many sessions do you mention in the Executive Summary?
   • How many OJ publications do you mention in the Executive Summary?
2. Count sources in the HTML Annex:
   • How many sessions listed in "Sources Consulted"?
   • How many OJ publications listed?
3. Compare with the INVENTORY from Phase 3.7:

   VALIDATION:
   - Summary says: X sessions, Y publications
   - Annex lists: A sessions, B publications
   - Inventory has: TOTAL_SESSIONS, TOTAL_OJ
   
   X == A == TOTAL_SESSIONS? → If NO, CORRECT
   Y == B == TOTAL_OJ? → If NO, CORRECT

4. If there's a discrepancy:
   • STOP
   • Correct the Executive Summary to use INVENTORY numbers
   • Make sure the Annex lists ALL sources from INVENTORY
   • Validate again
5. Validate LINKS:
   • Does each session in the Annex have

     <a href="https://parlamento.ai/[COUNTRY]/transcripts-full/[ID]">

     ?
   • Does each OJ document have

     <a href="PDF_URL">

     ?
   • Do quotes in the body have links to the original session?
   • If any link is missing, ADD IT before continuing
6. Only continue to Phase 6 when:

   ✓ Summary = Annex = Inventory (sessions)
   ✓ Summary = Annex = Inventory (OJ publications)
   ✓ ALL session IDs are clickable links
   ✓ ALL OJ documents have PDF links

1. Save the HTML using the Write tool (DO NOT use cat/heredoc):

Write("/tmp/report.html", your_complete_html)

1. Send the HTML directly to the PDF generator:

curl -X POST "https://source-worker-876875904047.us-central1.run.app/generate-pdf?filename=report.pdf" \
  -H "Authorization: Bearer $PARLAMENTO_API_KEY" \
  -H "Content-Type: text/html; charset=utf-8" \
  --data-binary "@/tmp/report.html" \
  --output "report.pdf"

1. Confirm to user:

✓ PDF generated: report.pdf

--data-binary

• Issue description
• Expressed positions
• Relevant quotes with link to session

<blockquote>
  "Desalination is not the future, it's the present..."
  <br><small>— Session of January 15, 2026
  (<a href="https://parlamento.ai/cl/transcripts-full/8645">view transcript</a>)</small>
</blockquote>

id

/transcript/[id]

• Observed patterns
• Discussion evolution
• Points of consensus and dissent

• Key points to remember
• Possible implications

1. DO NOT generate an empty PDF
2. Inform the user what you searched for
3. Suggest alternatives:
   • Other dates with available data
   • Similar committees
   • Expand the search range

1. List the found options
2. Ask the user which one they want
3. Continue with the selection

1. Include those that do have content
2. Mention in the report which sessions had limited information

1. Query

   /hierarchy?countryCode=cl

   → Find subgroupId=87 for "Health Committee"
2. Query

   /transcripts?countryCode=cl&subgroupIds=87&dateFrom=2026-01-13&status=completed

3. For each session, query

   /transcript/ID

   to get full content
4. READ all content
5. WRITE your analysis:
   • "The Health Committee met 2 times this week..."
   • "The main topics were: hospital reform, 2026 budget..."
   • "Representative X stated: '...relevant quote...'"
   • "A trend is observed towards..."
6. Generate the HTML with your analysis
7. Send to

   /generate-pdf

8. User receives professional PDF with your analysis

• DO NOT show raw data as markdown tables
• DO NOT copy and paste JSONs to the user
• DO NOT generate reports without having read the content
• DO NOT invent information that is not in the data
• DO NOT invent parliamentarian names - if the name doesn't appear in the transcript, don't include it
• DO NOT use generic attributions like "Project author representative" or "Regional senator" - if you don't know the name, omit the attribution or say "A parliamentarian stated..."
• DO NOT inflate metrics - if you read 40 sessions, report 40, not 89

1. Exact count (Triple verification):
   • The number in Executive Summary MUST = number in Annex MUST = Phase 3.7 INVENTORY
   • If you say "15 sessions were analyzed", the annex MUST list exactly 15 sessions
   • DO NOT round, DO NOT approximate, DO NOT invent numbers
2. Verifiable quotes:
   • Every quote MUST have: exact text + speaker name (if available) + session ID
   • The ID in the link MUST be real (from INVENTORY)
   • If you don't know the speaker's name, use "A parliamentarian stated..." but NEVER invent names
3. No inventions:
   • If you don't have the data, DO NOT invent it - omit it
   • If a session has no useful content, DO NOT count it in the analysis
   • A report with 5 real sources is better than one with 15 inflated sources
4. Complete annex:
   • The Annex MUST list ALL sources from INVENTORY
   • Each entry must have: ID, date, title/committee
   • DO NOT omit sources to "simplify"
5. MANDATORY LINKS (CRITICAL):
   • EVERY mentioned session MUST have clickable link to

     https://parlamento.ai/[COUNTRY]/transcripts-full/[ID]

   • EVERY Official Journal document MUST have link to original PDF
   • In the Annex, EACH entry must be a clickable

     <a href="URL">

   • DO NOT mention a source without its corresponding link
   • Transcript format:

     <a href="https://parlamento.ai/cl/transcripts-full/7677">Session 7677</a>

   • Official journal format:

     <a href="PDF_URL">View official document</a>