creation

workflows/Create a founder digest and leads from Hacker News with GPT-4o and Gmail-12499/readme-12499.md

@@ -0,0 +1,426 @@
+Create a founder digest and leads from Hacker News with GPT-4o and Gmail
+
+https://n8nworkflows.xyz/workflows/create-a-founder-digest-and-leads-from-hacker-news-with-gpt-4o-and-gmail-12499
+
+
+# Create a founder digest and leads from Hacker News with GPT-4o and Gmail
+
+## 1. Workflow Overview
+
+**Purpose:** This workflow runs daily to discover high-signal Hacker News threads (via Google SERP/SerpAPI), structure and filter them, then uses Azure OpenAI (GPT-4o-mini) to produce two outputs in parallel:
+1) **A daily “Founder Digest”** emailed via Gmail (trend + why it matters + actions).  
+2) **Founder/buying-intent insights (“leads/signals”)** extracted with an AI agent (optionally enriched via an MCP social tool) and appended to Google Sheets.
+
+**Target use cases:**
+- Founder/operator daily intelligence on HN (Show HN, Launch HN, AI, SaaS, startup signals).
+- Lightweight lead discovery and “market signal” logging for follow-up.
+
+### 1.1 Trigger & Discovery (SERP)
+Daily schedule → query Google results for HN discussion pages.
+
+### 1.2 Parsing & Filtering (Discussion Records)
+Convert SERP response into normalized discussion items → filter out non-thread/noise pages.
+
+### 1.3 Aggregation (Single dataset for AI)
+Aggregate all valid threads into a single object for downstream AI prompts.
+
+### 1.4 Output A: Founder Digest Email (AI + Gmail)
+AI agent generates a formatted digest → send email.
+
+### 1.5 Output B: Lead/Signal Extraction + Storage (AI + MCP + Sheets)
+AI agent extracts high-signal insights (with MCP tool available) → normalize text → append to Google Sheets.
+
+---
+
+## 2. Block-by-Block Analysis
+
+### Block 1.1 — Trigger & Search
+**Overview:** Runs on a schedule and fetches relevant Hacker News discussions using SerpAPI (Google SERP).  
+**Nodes involved:**  
+- Trigger Daily Hacker News Scan  
+- Search Hacker News Discussions via Google SERP
+
+#### Node: Trigger Daily Hacker News Scan
+- **Type / Role:** Schedule Trigger — time-based entrypoint.
+- **Configuration (interpreted):** Runs on an interval schedule (the JSON shows an interval rule placeholder; in the UI this should be set to “every day” at a chosen time).
+- **Connections:** Outputs to **Search Hacker News Discussions via Google SERP**.
+- **Failure/edge cases:**
+  - Misconfigured schedule (e.g., interval not set as intended) → workflow not firing.
+  - Instance timezone differences can shift expected delivery time.
+
+#### Node: Search Hacker News Discussions via Google SERP
+- **Type / Role:** SerpAPI node — executes a Google query.
+- **Configuration (interpreted):**
+  - Query: `site:news.ycombinator.com ("Show HN" OR "Launch HN" OR "AI" OR "startup" OR "SaaS")`
+  - Returns **5** results (`num: 5`).
+- **Input:** Trigger.
+- **Output:** Raw SerpAPI response to **Extract SERP Results into Discussion Records**.
+- **Credentials:** Requires SerpAPI API key/credentials in n8n.
+- **Failure/edge cases:**
+  - SerpAPI auth errors / quota exhaustion.
+  - Google SERP variability: results can include non-thread pages, duplicates, or irrelevant pages.
+  - Small `num=5` can miss important threads on busy days.
+
+**Sticky note (applies to nodes in this block):**  
+“## 📡 Trigger & Search … Scheduled daily scan … Searches for Show HN, Launch HN, AI, startup, and SaaS …”
+
+---
+
+### Block 1.2 — Processing: Extract and Filter Discussion Records
+**Overview:** Transforms the SERP response into clean per-thread items, then filters out known noise titles and ensures links are actual discussion threads.  
+**Nodes involved:**  
+- Extract SERP Results into Discussion Records  
+- Filter Valid Hacker News Discussion Threads
+
+#### Node: Extract SERP Results into Discussion Records
+- **Type / Role:** Code node — maps SerpAPI’s `organic_results` into a simplified schema.
+- **Configuration (interpreted):**
+  - Reads first input item: `const data = $input.first().json;`
+  - Uses `data.organic_results || []`
+  - Outputs items with fields: `title`, `summary`(snippet), `link`, `date`, `source`
+- **Input:** SerpAPI response.
+- **Output:** List of structured items to **Filter Valid Hacker News Discussion Threads**.
+- **Failure/edge cases:**
+  - SerpAPI response shape changes (missing `organic_results`) → outputs empty list.
+  - Missing fields: node defaults to empty strings and “Recent”.
+  - If SerpAPI returns fewer than 5 results, downstream aggregation may be sparse.
+
+#### Node: Filter Valid Hacker News Discussion Threads
+- **Type / Role:** Code node — removes irrelevant/noise pages and keeps only real HN threads.
+- **Configuration (interpreted):**
+  - Blocks specific titles:
+    - “Show | Hacker News”
+    - “Show HN Guidelines”
+    - “Launch HNs”
+    - “Launch HN Instructions”
+  - Keeps only links containing: `news.ycombinator.com/item`
+  - Re-outputs normalized fields (title/summary/link/date/source)
+- **Input:** Structured SERP items.
+- **Output:** Filtered discussion items to **Aggregate Hacker News Discussions for Analysis**.
+- **Failure/edge cases:**
+  - Legitimate threads might be excluded if titles match blocked titles unexpectedly.
+  - Threads could be missed if URL format differs (mobile URLs, tracking parameters, alternate HN domains).
+  - If everything is filtered out → downstream AI gets an empty dataset.
+
+**Sticky note (applies to nodes in this block):**  
+“## 🔍 Data Processing … Extracts … filters out noise pages … keeping only real discussion threads …”
+
+---
+
+### Block 1.3 — Aggregation (Prepare dataset for AI)
+**Overview:** Combines multiple filtered thread items into one aggregated dataset so AI nodes can reason over the full set for the day.  
+**Nodes involved:**  
+- Aggregate Hacker News Discussions for Analysis
+
+#### Node: Aggregate Hacker News Discussions for Analysis
+- **Type / Role:** Aggregate node — aggregates all incoming item JSON into one item.
+- **Configuration (interpreted):**
+  - Mode: “Aggregate All Item Data” (collects all items into a single array, typically under `data` in output).
+- **Input:** Filtered thread items.
+- **Output connections:** Splits to two parallel branches:
+  - **Generate Founder Digest from Hacker News Discussions (AI)**
+  - **Extract High-Signal Founder & Buying-Intent Insights (AI)**
+- **Failure/edge cases:**
+  - Empty input → aggregated `data` may be `[]`; AI prompts will still run unless you add an IF guard.
+  - Large inputs (if `num` increased significantly) can enlarge prompt size and hit model context limits.
+
+**Sticky note (applies to this block):**  
+“## 📦 Discussion Aggregation … Combines all valid … into a single dataset …”
+
+---
+
+### Block 1.4 — Founder Digest Generation & Email Delivery
+**Overview:** Uses an AI agent (GPT-4o-mini on Azure OpenAI) to generate a strict-format digest and emails it via Gmail.  
+**Nodes involved:**  
+- LLM Reasoning Engine for Founder Digest  
+- Generate Founder Digest from Hacker News Discussions (AI)  
+- Email Founder Daily Hacker News Digest
+
+#### Node: LLM Reasoning Engine for Founder Digest
+- **Type / Role:** Azure OpenAI Chat Model (LangChain) — provides the language model for the agent.
+- **Configuration (interpreted):**
+  - Model: `gpt-4o-mini`
+- **Connections:** Connected to the agent via `ai_languageModel`.
+- **Credentials:** Azure OpenAI credentials required (endpoint, API key, deployment/model mapping depending on n8n setup).
+- **Failure/edge cases:**
+  - Azure auth/endpoint misconfiguration.
+  - Model/deployment name mismatch.
+  - Rate limits/timeouts during scheduled runs.
+
+#### Node: Generate Founder Digest from Hacker News Discussions (AI)
+- **Type / Role:** LangChain Agent — generates the digest text using provided system + user prompt.
+- **Configuration (interpreted):**
+  - Prompt injects aggregated data: `{{ JSON.stringify($json.data, null, 2) }}`
+  - Enforces exact format:
+    - “🔥 Hacker News Daily Digest”
+    - “📌 Key Trend”
+    - “🧠 Why It Matters”
+    - “🚀 Founder Actions”
+  - System message: “startup and technology trend analyst… avoid fluff… concise and actionable”
+- **Inputs:**
+  - Main input: aggregated dataset (expects `$json.data`).
+  - AI language model input: from **LLM Reasoning Engine for Founder Digest**.
+- **Output:** Typically agent output is accessible as `{{$json.output}}` for the next node (as used by Gmail).
+- **Failure/edge cases:**
+  - If `$json.data` is undefined (aggregation changed or empty) → prompt may degrade or error.
+  - Output may deviate from “format EXACTLY” in edge cases; consider adding post-validation.
+
+#### Node: Email Founder Daily Hacker News Digest
+- **Type / Role:** Gmail node — sends the digest email.
+- **Configuration (interpreted):**
+  - To: `user@example.com` (placeholder; must be replaced)
+  - Subject: `🔥 Hacker News Daily Digest`
+  - Body: `={{ $json.output }}`
+- **Input:** AI agent output.
+- **Credentials:** Gmail OAuth2.
+- **Failure/edge cases:**
+  - OAuth2 token expiry/refresh issues.
+  - Gmail sending limits, “from” restrictions, or blocked content rules.
+  - If AI output is empty → sends blank email unless guarded.
+
+**Sticky note (applies to nodes in this block):**  
+“## 📧 Founder Digest … daily email digest … Uses AI to synthesize discussions …”
+
+---
+
+### Block 1.5 — Lead Discovery / Signal Extraction + Enrichment + Storage
+**Overview:** Uses an AI agent to extract buying-intent/founder signals from the aggregated HN dataset. The agent has an MCP tool available for external enrichment. Output is normalized and appended to Google Sheets.  
+**Nodes involved:**  
+- LLM Engine for Lead Qualification & Signal Interpretation  
+- External Social Context & Enrichment Tool (MCP Client)  
+- Extract High-Signal Founder & Buying-Intent Insights (AI)  
+- Normalize AI Insight Output for Storage  
+- Append Founder Insights to Hacker News Intelligence Sheet
+
+#### Node: LLM Engine for Lead Qualification & Signal Interpretation
+- **Type / Role:** Azure OpenAI Chat Model (LangChain) — provides the model for the lead/signal agent.
+- **Configuration (interpreted):**
+  - Model: `gpt-4o-mini`
+- **Connections:** Feeds the agent via `ai_languageModel`.
+- **Credentials:** Azure OpenAI.
+- **Failure/edge cases:** Same as other Azure model node (auth, rate limits, deployment mismatch).
+
+#### Node: External Social Context & Enrichment Tool (MCP Client)
+- **Type / Role:** MCP Client Tool (LangChain tool) — optional enrichment tool the agent can call.
+- **Configuration (interpreted):**
+  - Endpoint: `https://mcp.xpoz.ai/mcp`
+  - Auth: Bearer token auth (`bearerAuth`)
+- **Connections:** Exposed to the agent via `ai_tool`.
+- **Credentials:** Requires MCP bearer token/credential in n8n.
+- **Failure/edge cases:**
+  - Endpoint downtime / TLS issues.
+  - Auth token invalid/expired.
+  - Tool results may be non-deterministic; agent may overuse tool calls unless constrained.
+
+#### Node: Extract High-Signal Founder & Buying-Intent Insights (AI)
+- **Type / Role:** LangChain Agent — extracts structured “founder intelligence” from the aggregated dataset.
+- **Configuration (interpreted):**
+  - Prompt instructs:
+    1) Key Signal  
+    2) Problem Statement  
+    3) Why This Matters  
+    4) Actionable Insight  
+  - Constraints: “Use ONLY the provided data. Do NOT invent facts.”
+  - System message emphasizes high-signal insights and ignoring generic content.
+  - Max iterations: 30 (agent may loop/tool-call up to this limit).
+  - Injects `{{ JSON.stringify($json.data, null, 2) }}`
+- **Inputs:**
+  - Main: aggregated dataset from Aggregate node.
+  - AI language model: **LLM Engine for Lead Qualification & Signal Interpretation**.
+  - Tool: **External Social Context & Enrichment Tool (MCP Client)**.
+- **Output:** Passed to normalization node; expects `item.json.output` to exist.
+- **Failure/edge cases:**
+  - Empty `$json.data` → weak/empty insights.
+  - “Do not invent facts” can conflict with enrichment tool usage unless the agent clearly attributes tool output vs. HN data.
+  - Long tool/agent loops can hit execution time limits.
+
+#### Node: Normalize AI Insight Output for Storage
+- **Type / Role:** Code node — cleans AI output formatting for easier Sheet storage.
+- **Configuration (interpreted):**
+  - Reads `item.json.output`
+  - Removes numbered bold headings via regex: `/\d+\.\s*\*\*.*?\*\*\n?/g`
+  - Removes bullet markers `- `
+  - Collapses extra line breaks, trims whitespace
+  - Outputs `{ relaxed_output: text }`
+- **Input:** AI agent output.
+- **Output:** Clean text into **Append Founder Insights to Hacker News Intelligence Sheet**.
+- **Failure/edge cases:**
+  - If the agent output format changes, regex may remove unintended parts or not remove headings.
+  - If `output` is missing → stores empty string.
+
+#### Node: Append Founder Insights to Hacker News Intelligence Sheet
+- **Type / Role:** Google Sheets node — appends a row with the normalized insight.
+- **Configuration (interpreted):**
+  - Operation: Append
+  - Document: Spreadsheet ID `YourSpreadsheetId` (placeholder)
+  - Sheet/tab name: `hacker_news`
+  - Column mapping: writes `output = {{$json.relaxed_output}}`
+  - Matching columns includes `output` (not typically needed for append; but configured)
+- **Input:** Normalized `{ relaxed_output }`.
+- **Credentials:** Google Sheets OAuth2.
+- **Failure/edge cases:**
+  - Spreadsheet/tab not found or permissions missing.
+  - Column schema mismatch (if the sheet doesn’t have an `output` column).
+  - Google API quota/rate limits.
+
+**Sticky note (applies to nodes in this block):**  
+“## 🎯 Lead Discovery … identify buying-intent signals … enrich … log to Google Sheets …”
+
+---
+
+## 3. Summary Table
+
+| Node Name | Node Type | Functional Role | Input Node(s) | Output Node(s) | Sticky Note |
+|---|---|---|---|---|---|
+| Main Overview | Sticky Note | Documentation / canvas overview | — | — | “## 🚀 Automate Founder Digest & Lead Discovery… Setup steps… Connect SerpAPI… Azure OpenAI… Gmail… Google Sheets… MCP…” |
+| Section: Trigger | Sticky Note | Documentation for trigger/search area | — | — | “## 📡 Trigger & Search… Scheduled daily scan… SERP API…” |
+| Section: Processing | Sticky Note | Documentation for parsing/filtering | — | — | “## 🔍 Data Processing… filters out noise pages…” |
+| Section: Email Digest | Sticky Note | Documentation for digest branch | — | — | “## 📧 Founder Digest… daily email digest…” |
+| Section: Lead Discovery | Sticky Note | Documentation for lead branch | — | — | “## 🎯 Lead Discovery… enrich… log to Google Sheets…” |
+| Security Note | Sticky Note | Credential requirements note | — | — | “## 🔐 Required Credentials… SerpAPI… Azure OpenAI… Gmail… Google Sheets… Xpoz MCP…” |
+| Trigger Daily Hacker News Scan | Schedule Trigger | Entry point (daily run) | — | Search Hacker News Discussions via Google SERP | “## 📡 Trigger & Search… Scheduled daily scan…” |
+| Search Hacker News Discussions via Google SERP | SerpAPI | Google SERP discovery for HN | Trigger Daily Hacker News Scan | Extract SERP Results into Discussion Records | “## 📡 Trigger & Search… Scheduled daily scan…” |
+| Extract SERP Results into Discussion Records | Code | Parse SERP response to discussion items | Search Hacker News Discussions via Google SERP | Filter Valid Hacker News Discussion Threads | “## 🔍 Data Processing… filters out noise pages…” |
+| Filter Valid Hacker News Discussion Threads | Code | Keep only real HN `/item?id=` threads | Extract SERP Results into Discussion Records | Aggregate Hacker News Discussions for Analysis | “## 🔍 Data Processing… filters out noise pages…” |
+| Sticky Note | Sticky Note | Notes on aggregation | — | — | “## 📦 Discussion Aggregation… Combines all valid… into a single dataset…” |
+| Aggregate Hacker News Discussions for Analysis | Aggregate | Combine all threads into `$json.data` | Filter Valid Hacker News Discussion Threads | Generate Founder Digest from Hacker News Discussions (AI); Extract High-Signal Founder & Buying-Intent Insights (AI) | “## 📦 Discussion Aggregation… Combines all valid… into a single dataset…” |
+| LLM Reasoning Engine for Founder Digest | Azure OpenAI Chat Model (LangChain) | LLM provider for digest agent | — (model connection) | Generate Founder Digest from Hacker News Discussions (AI) | “## 📧 Founder Digest… daily email digest…” |
+| Generate Founder Digest from Hacker News Discussions (AI) | LangChain Agent | Create formatted founder digest | Aggregate Hacker News Discussions for Analysis (+ model) | Email Founder Daily Hacker News Digest | “## 📧 Founder Digest… daily email digest…” |
+| Email Founder Daily Hacker News Digest | Gmail | Email the digest | Generate Founder Digest from Hacker News Discussions (AI) | — | “## 📧 Founder Digest… daily email digest…” |
+| LLM Engine for Lead Qualification & Signal Interpretation | Azure OpenAI Chat Model (LangChain) | LLM provider for lead agent | — (model connection) | Extract High-Signal Founder & Buying-Intent Insights (AI) | “## 🎯 Lead Discovery… enrich… log to Google Sheets…” |
+| External Social Context & Enrichment Tool (MCP Client) | MCP Client Tool (LangChain) | Optional social enrichment tool | — (tool connection) | Extract High-Signal Founder & Buying-Intent Insights (AI) | “## 🎯 Lead Discovery… enrich… log to Google Sheets…” |
+| Extract High-Signal Founder & Buying-Intent Insights (AI) | LangChain Agent | Extract buying-intent/founder signals | Aggregate Hacker News Discussions for Analysis (+ model + tool) | Normalize AI Insight Output for Storage | “## 🎯 Lead Discovery… enrich… log to Google Sheets…” |
+| Normalize AI Insight Output for Storage | Code | Clean agent output for Sheets | Extract High-Signal Founder & Buying-Intent Insights (AI) | Append Founder Insights to Hacker News Intelligence Sheet | “## 🎯 Lead Discovery… enrich… log to Google Sheets…” |
+| Append Founder Insights to Hacker News Intelligence Sheet | Google Sheets | Append insights to spreadsheet | Normalize AI Insight Output for Storage | — | “## 🎯 Lead Discovery… enrich… log to Google Sheets…” |
+
+---
+
+## 4. Reproducing the Workflow from Scratch
+
+1) **Create a new workflow**
+- Name it: *Automate Founder Digest & Lead Discovery from Hacker News with GPT-4o* (or your preferred name).
+- Ensure execution order is default (v1 is fine).
+
+2) **Add Schedule Trigger**
+- Add node: **Schedule Trigger**
+- Configure to run **daily** at your desired time (match your timezone).
+- Rename: **Trigger Daily Hacker News Scan**
+
+3) **Add SerpAPI search node**
+- Add node: **SerpAPI** (SerpAPI integration)
+- Rename: **Search Hacker News Discussions via Google SERP**
+- Credentials: connect your **SerpAPI** credential.
+- Set query `q` to:  
+  `site:news.ycombinator.com ("Show HN" OR "Launch HN" OR "AI" OR "startup" OR "SaaS")`
+- Set number of results (`num`) to **5** (or adjust).
+- Connect: **Schedule Trigger → SerpAPI**
+
+4) **Add Code node to extract results**
+- Add node: **Code**
+- Rename: **Extract SERP Results into Discussion Records**
+- Paste logic that:
+  - Reads `$input.first().json.organic_results || []`
+  - Maps to `{ title, summary: snippet, link, date, source }`
+  - Returns one n8n item per result
+- Connect: **SerpAPI → Extract Code**
+
+5) **Add Code node to filter valid HN discussion threads**
+- Add node: **Code**
+- Rename: **Filter Valid Hacker News Discussion Threads**
+- Implement filtering:
+  - Remove noise titles (guidelines / Show page)
+  - Keep only URLs containing `news.ycombinator.com/item`
+- Connect: **Extract Code → Filter Code**
+
+6) **Add Aggregate node**
+- Add node: **Aggregate**
+- Rename: **Aggregate Hacker News Discussions for Analysis**
+- Set aggregation mode to **Aggregate All Item Data** (single output item containing an array of all items, typically under `data`).
+- Connect: **Filter Code → Aggregate**
+
+7) **Set up Azure OpenAI Chat Model (digest)**
+- Add node: **Azure OpenAI Chat Model** (LangChain)
+- Rename: **LLM Reasoning Engine for Founder Digest**
+- Choose model/deployment: **gpt-4o-mini**
+- Configure Azure OpenAI credentials (endpoint, API key, deployment/model mapping as required by your n8n version).
+
+8) **Add LangChain Agent for founder digest**
+- Add node: **AI Agent** (LangChain Agent)
+- Rename: **Generate Founder Digest from Hacker News Discussions (AI)**
+- Prompt type: “Define”
+- System message: trend analyst, concise, actionable.
+- User prompt:
+  - Include strict output format headings
+  - Include `{{ JSON.stringify($json.data, null, 2) }}` as the input dataset
+- Connect main: **Aggregate → Digest Agent**
+- Connect AI language model: **LLM Reasoning Engine for Founder Digest → Digest Agent** (language model connection)
+
+9) **Add Gmail send node**
+- Add node: **Gmail** → operation **Send**
+- Rename: **Email Founder Daily Hacker News Digest**
+- Credentials: Gmail OAuth2 (connect account).
+- To: your email address (replace `user@example.com`)
+- Subject: `🔥 Hacker News Daily Digest`
+- Message/body: set expression to `{{ $json.output }}`
+- Connect: **Digest Agent → Gmail**
+
+10) **Set up Azure OpenAI Chat Model (lead/signal)**
+- Add node: **Azure OpenAI Chat Model** (LangChain)
+- Rename: **LLM Engine for Lead Qualification & Signal Interpretation**
+- Model/deployment: **gpt-4o-mini**
+- Same Azure credentials (or separate, if desired).
+
+11) **Add MCP Client Tool node**
+- Add node: **MCP Client Tool** (LangChain)
+- Rename: **External Social Context & Enrichment Tool (MCP Client)**
+- Endpoint URL: `https://mcp.xpoz.ai/mcp`
+- Auth: **Bearer**
+- Credentials: store bearer token securely in n8n credentials.
+
+12) **Add LangChain Agent for lead/signal extraction**
+- Add node: **AI Agent** (LangChain Agent)
+- Rename: **Extract High-Signal Founder & Buying-Intent Insights (AI)**
+- Prompt type: “Define”
+- Set max iterations to **30**
+- System message: founder intelligence analyst; high-signal only; no fluff.
+- User prompt:
+  - Ask for: Key Signal, Problem Statement, Why This Matters, Actionable Insight
+  - Add constraint “use only provided data; do not invent facts”
+  - Inject dataset via `{{ JSON.stringify($json.data, null, 2) }}`
+- Connect main: **Aggregate → Lead Agent**
+- Connect AI language model: **LLM Engine… → Lead Agent**
+- Connect tool: **MCP Client Tool → Lead Agent** (tool connection)
+
+13) **Add Code node to normalize output for storage**
+- Add node: **Code**
+- Rename: **Normalize AI Insight Output for Storage**
+- Logic:
+  - Read `item.json.output`
+  - Strip list markers / headings
+  - Output `{ relaxed_output: text }`
+- Connect: **Lead Agent → Normalize Code**
+
+14) **Add Google Sheets append node**
+- Add node: **Google Sheets**
+- Rename: **Append Founder Insights to Hacker News Intelligence Sheet**
+- Credentials: Google Sheets OAuth2.
+- Operation: **Append**
+- Spreadsheet: select your target spreadsheet (replace `YourSpreadsheetId`)
+- Sheet/tab: `hacker_news` (create it if missing)
+- Ensure a column named `output` exists, then map:
+  - `output` = `{{ $json.relaxed_output }}`
+- Connect: **Normalize Code → Google Sheets**
+
+15) **(Optional) Add guardrails**
+- Add an **IF** node after filtering or after aggregation:
+  - If `data` is empty, skip AI + email/sheets to avoid noisy empty outputs.
+
+---
+
+## 5. General Notes & Resources
+
+| Note Content | Context or Link |
+|---|---|
+| Required credentials: SerpAPI, Azure OpenAI, Gmail OAuth2, Google Sheets OAuth2, Xpoz MCP (bearer). Use OAuth2 where possible and never share personal API keys. | Security note (in-workflow sticky note) |
+| This workflow turns Hacker News into a daily intelligence engine by combining SERP-based discovery, AI reasoning, and structured storage. | Main overview sticky note |
+| Disclaimer (provided by user): Le texte fourni provient exclusivement d’un workflow automatisé réalisé avec n8n… | User-provided disclaimer/context |