Niche Deep-Dive

Use these workflows to progressively research a niche and generate content hooks. Each level goes deeper. Enter at any level, chain them together, or use the one-shot shortcut.

Progression Overview

Level 1: Landscape        global/trend-research           "What's trending?"
         │
Level 2: Topics           research/discover-topics        "What specific topics are hot?"
         │
Level 3: Deep Research    global/deep-research            "What does the internet say?"
         │
Level 4: Problems         research/problem-intelligence   "What pain points exist?"
         │
Level 5: Hooks            hooks/generate                  "Give me viral hooks"

Shortcut ─────────────── global/research-to-hooks ─────── "Research + hooks in one call"

Level	Workflow	Required Input	Uses Niche?
1	global/trend-research	none (defaults to “technology”)	Yes
2	research/discover-topics	niche	Yes (required)
3	global/deep-research	query	No (uses query)
4	research/problem-intelligence	one of: niche / query / subreddits	Yes (optional)
5	hooks/generate	niche	Yes (required)
Shortcut	global/research-to-hooks	query	Derived

---

Level 1: Landscape

Workflow: global/trend-research Question: What’s trending in this niche right now?

Start here when a user picks a niche and you want a broad view of what’s hot.

REST

curl -X POST 'https://gofabric.dev/v1/workflows/run?name=global/trend-research&organization_id=<org-id>' \
  -H 'Authorization: Bearer fab_xxx' \
  -H 'Content-Type: application/json' \
  -d '{
    "input": {
      "niche": "AI tools",
      "platform": "YouTube,TikTok,Instagram",
      "time_range": "last 7 days"
    }
  }'

GraphQL

mutation {
  submitWorkflowRun(
    name: "global/trend-research"
    orgId: "<org-id>"
    input: {
      niche: "AI tools"
      platform: "YouTube,TikTok,Instagram"
      time_range: "last 7 days"
    }
  ) {
    id
    status
  }
}

All fields optional — defaults: niche="technology", platform="YouTube,TikTok,Instagram", time_range="last 7 days".

Output

{
  "trends": {
    "trending_topics": [
      { "topic": "AI Code Assistants", "virality_score": 9, "growth_direction": "rising" }
    ],
    "viral_formats": [
      { "format": "Before/After demos", "why_it_works": "Visual proof of value" }
    ],
    "common_hooks": ["Stop using ChatGPT wrong", "This AI tool replaced my..."],
    "emerging_angles": ["AI tool fatigue", "Open-source alternatives"],
    "trending_hashtags": ["#AItools", "#DevProductivity"]
  },
  "content_ideas": [
    { "title": "5 AI tools nobody talks about", "hook": "...", "format": "short", "engagement": 8 }
  ]
}

Next step: Show trending_topics and content_ideas to the user. Let them pick a topic → feed it into Level 2 or 3. Save trends to enrich Level 5.

---

Level 2: Topics

Workflow: research/discover-topics Question: What specific topics are hot and underserved?

Use when a user has a niche and wants to find the best specific topic to create content about.

REST

curl -X POST 'https://gofabric.dev/v1/workflows/run?name=research/discover-topics&organization_id=<org-id>' \
  -H 'Authorization: Bearer fab_xxx' \
  -H 'Content-Type: application/json' \
  -d '{
    "input": {
      "niche": "AI tools",
      "num_topics": 10,
      "min_score": 5
    }
  }'

GraphQL

mutation {
  submitWorkflowRun(
    name: "research/discover-topics"
    orgId: "<org-id>"
    input: { niche: "AI tools", num_topics: 10, min_score: 5 }
  ) {
    id
    status
  }
}

niche is required. num_topics defaults to 10, min_score defaults to 5 (1–10 scale).

Output

{
  "topics": [
    {
      "topic": "Cursor vs Copilot in 2026",
      "headline": "The AI editor war nobody expected",
      "virality_score": 8,
      "content_angle": "Contrarian take — Copilot is catching up",
      "why_trending": "Cursor pricing backlash + Copilot Workspace launch",
      "competition": "medium",
      "timeliness": "this_week"
    }
  ],
  "best_topic": { "topic": "...", "virality_score": 9 },
  "discovery_context": "formatted research text for downstream stages"
}

Next step: Show ranked topics. Use best_topic.topic (or user’s pick) as the query for Level 3.

---

Level 3: Deep Research

Workflow: global/deep-research Question: What does the internet say about this topic?

Multi-source parallel research across web, YouTube, Reddit, and RSS.

REST

curl -X POST 'https://gofabric.dev/v1/workflows/run?name=global/deep-research&organization_id=<org-id>' \
  -H 'Authorization: Bearer fab_xxx' \
  -H 'Content-Type: application/json' \
  -d '{
    "input": {
      "query": "AI code generation tools for startups",
      "depth": 5
    }
  }'

GraphQL

mutation {
  submitWorkflowRun(
    name: "global/deep-research"
    orgId: "<org-id>"
    input: { query: "AI code generation tools for startups", depth: 5 }
  ) {
    id
    status
  }
}

query is required. depth is optional: 1–3 = quick, 5 = standard, 10 = deep. Also accepts sources (list) and rss_feeds (string).

Output

{
  "synthesis": {
    "key_findings": ["Cursor dominates mindshare but pricing is a pain point", "..."],
    "themes": ["Open-source alternatives gaining traction", "Enterprise adoption accelerating"],
    "notable_sources": [{ "title": "...", "url": "...", "why": "..." }],
    "confidence": 0.85,
    "gaps": ["Limited data on enterprise pricing comparisons"]
  },
  "raw_sources": {
    "web": [{ "title": "...", "url": "...", "snippet": "..." }],
    "youtube": [{ "title": "...", "channel": "...", "views": 150000 }],
    "reddit": [{ "title": "...", "subreddit": "programming" }]
  },
  "sources_queried": ["web", "youtube", "reddit"],
  "total_sources_collected": 42
}

Next step: Show synthesis as a research summary. Feed the full output into global/research-to-hooks, or extract themes and raw_sources to enrich Level 5.

---

Level 4: Problem Intelligence

Workflow: research/problem-intelligence Question: What pain points and opportunities exist in this niche?

Deepest research — ingests from Reddit, HackerNews, YouTube, TikTok, Instagram, Twitter, and forums. Clusters problems, scores opportunities, generates ideas.

REST

curl -X POST 'https://gofabric.dev/v1/workflows/run?name=research/problem-intelligence&organization_id=<org-id>' \
  -H 'Authorization: Bearer fab_xxx' \
  -H 'Content-Type: application/json' \
  -d '{
    "input": {
      "niche": "AI tools",
      "depth": 10,
      "platforms": ["reddit", "hackernews", "youtube"]
    }
  }'

GraphQL

mutation {
  submitWorkflowRun(
    name: "research/problem-intelligence"
    orgId: "<org-id>"
    input: {
      niche: "AI tools"
      depth: 10
      platforms: ["reddit", "hackernews", "youtube"]
    }
  ) {
    id
    status
  }
}

Requires at least one of: niche, query, subreddits, or forum_urls. Depth: 3 = quick, 10 = standard, 25 = deep, 50 = exhaustive. Set reddit_discovery: true to auto-discover relevant subreddits.

See Research Workflows for the full input reference and scoring formula.

Output

{
  "ranked_clusters": [
    {
      "cluster_id": "cluster_0",
      "canonical_problem": "AI tool pricing is unpredictable — users hit surprise bills",
      "mention_count": 45,
      "opportunity_score": 0.87,
      "sentiment": "negative",
      "frustration_level": 8
    }
  ],
  "ideas": [
    {
      "name": "CostGuard AI",
      "core_value_proposition": "Predictable AI tool billing with usage caps",
      "viability_score": 8.5,
      "problem_cluster_id": "cluster_0"
    }
  ],
  "platform_summary": { "reddit": 60, "hackernews": 40, "youtube": 20 },
  "total_sources_collected": 150
}

Next step: Use ranked_clusters as content angles. Feed pain points into Level 5 via context_pack.pain_points.

---

Level 5: Hooks

Workflow: hooks/generate Question: Give me viral hooks grounded in all the intel.

Works standalone with just a niche, but produces the best results when enriched with data from Levels 1–4.

See Hook Generation for the full hook type taxonomy, emotional triggers, and 70/20/10 strategy.

Minimal input

REST

curl -X POST 'https://gofabric.dev/v1/workflows/run?name=hooks/generate&organization_id=<org-id>' \
  -H 'Authorization: Bearer fab_xxx' \
  -H 'Content-Type: application/json' \
  -d '{
    "input": {
      "niche": "AI tools",
      "platform": "TikTok",
      "num_hooks": 15
    }
  }'

GraphQL

mutation {
  submitWorkflowRun(
    name: "hooks/generate"
    orgId: "<org-id>"
    input: { niche: "AI tools", platform: "TikTok", num_hooks: 15 }
  ) {
    id
    status
  }
}

Enriched input (best results)

Pass data from earlier levels to ground hooks in real intelligence:

{
  "input": {
    "niche": "AI tools",
    "platform": "TikTok",
    "num_hooks": 15,
    "strategy": {
      "pillars": ["tutorials", "reviews", "hot takes"],
      "tone": "casual and edgy",
      "brand_voice": "tech-savvy friend"
    },
    "context_pack": {
      "pain_points": ["AI tool pricing is unpredictable"],
      "desires": ["one tool that does everything"],
      "target_audience": "startup founders"
    },
    "trends": {
      "trending_topics": [{ "topic": "AI Code Assistants", "virality_score": 9 }],
      "common_hooks": ["Stop using ChatGPT wrong"]
    }
  }
}

Field	Source
trends	Level 1 output .trends
context_pack.pain_points	Level 4 output .ranked_clusters[].canonical_problem
competitor_insights	hooks/scrape output or Level 3 .raw_sources
strategy	Workspace settings in Socialite

Output

{
  "hook_ideas": [
    {
      "hook_text": "What if I told you your IDE is lying to you?",
      "hook_type": "contrarian",
      "emotional_trigger": "curiosity",
      "tone": "provocative",
      "cta_text": "Follow for more AI hacks",
      "cta_type": "follow",
      "rationale": "Contrarian hooks on familiar tools drive debate in comments",
      "inspiration_source": "exploit",
      "estimated_strength": 9
    }
  ],
  "strategy_split": { "exploit": 10, "explore": 3, "wild": 2 }
}

---

One-Shot Shortcut

Workflow: global/research-to-hooks Question: Research this topic AND give me hooks — one call.

Chains global/deep-research → auto-bridge → hooks/generate. The bridge maps research synthesis.themes → trend signals and raw_sources → competitor insights automatically.

REST

curl -X POST 'https://gofabric.dev/v1/workflows/run?name=global/research-to-hooks&organization_id=<org-id>' \
  -H 'Authorization: Bearer fab_xxx' \
  -H 'Content-Type: application/json' \
  -d '{ "input": { "query": "AI code generation tools" } }'

GraphQL

mutation {
  submitWorkflowRun(
    name: "global/research-to-hooks"
    orgId: "<org-id>"
    input: { query: "AI code generation tools" }
  ) {
    id
    status
  }
}

Output combines the full deep research output with hook_ideas.

---

Optional Enrichment

Scrape competitor hooks

Workflow: hooks/scrape

Scrape real hooks from creator channels to use as competitor_insights in Level 5.

{
  "input": {
    "channels": ["@fireship", "@techlead"],
    "platform": "youtube",
    "max_per_channel": 30,
    "shorts_only": true
  }
}

Returns hooks with engagement rates. Feed the output into hooks/generate as competitor_insights.

Data-driven hook generation

Workflow: hooks/generate-from-data

Expects output from hooks/scrape → extract → cluster piped in. Use when you have scraped competitor data and want hooks grounded entirely in that data.

---

Chaining Outputs

Level 1 (trends)
  └─ .trends.common_hooks ─────────────────────────┐
  └─ .trends.trending_topics → user picks topic     │
                                    │                │
Level 2 (discover-topics)           │                │
  └─ .best_topic.topic ───────────┐ │               │
                                   │ │               │
Level 3 (deep-research)  ← query ─┘ │               │
  └─ .synthesis.themes ─────────────│───────────────┐│
  └─ .raw_sources ──────────────────│──────────────┐││
                                    │              │││
Level 4 (problem-intelligence)      │              │││
  └─ .ranked_clusters[].pain ──────┐│              │││
                                    ││             │││
Level 5 (hooks/generate) ──────────┘┘             │││
  ├─ context_pack.pain_points ← Level 4          │││
  ├─ trends                   ← Level 1 ─────────┘││
  ├─ competitor_insights      ← scrape / Level 3 ──┘│
  └─ strategy                 ← workspace settings   │

Note

You don’t need to run every level. The most common patterns are:

• Quick hooks: Level 5 alone (niche only, no enrichment)
• Informed hooks: Level 1 → Level 5 (trends → hooks)
• Research-first: Level 3 → Level 5 (or use the global/research-to-hooks shortcut)
• Full funnel: Level 1 → Level 2 → Level 3 → Level 4 → Level 5