Sandcastle/sandcastle.yaml.example at main · gizmax/Sandcastle · GitHub

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
name: lead-enrichment
description: Enrich companies with firmographic data

# Global settings
sandstorm_url: ${SANDSTORM_URL}
default_model: sonnet
default_max_turns: 10
default_timeout: 300

steps:
  - id: scrape
    prompt: |
      Scrape the website {input.website} for company {input.name}.
      Extract: company description, approximate team size,
      tech stack, pricing model, headquarters location.
      Return ONLY valid JSON matching the output schema.
    model: sonnet
    max_turns: 10
    timeout: 120
    parallel_over: input.companies
    output_schema:
      type: object
      properties:
        description: { type: string }
        employee_estimate: { type: string }
        tech_stack: { type: array, items: { type: string } }
        pricing_model: { type: string }
        hq_location: { type: string }
    retry:
      max_attempts: 3
      backoff: exponential
      on_failure: skip

  - id: enrich
    depends_on: [scrape]
    prompt: |
      Company: {input.name} ({input.website})
      Scraped data: {steps.scrape.output}
      Find additional information using web search:
      - Total funding and last funding round
      - Key contacts: CEO, CTO, VP Sales with LinkedIn URLs
      - Recent news from the last 6 months
      Return ONLY valid JSON.
    model: sonnet
    max_turns: 15
    output_schema:
      type: object
      properties:
        funding_total_usd: { type: number }
        last_round: { type: string }
        contacts:
          type: array
          items:
            type: object
            properties:
              name: { type: string }
              title: { type: string }
              linkedin_url: { type: string }
        recent_news:
          type: array
          items:
            type: object
            properties:
              headline: { type: string }
              date: { type: string }
              url: { type: string }

  - id: score
    depends_on: [scrape, enrich]
    prompt: |
      Score this lead 1-100 for a B2B SaaS sales team.
      Company: {input.name}
      Scrape data: {steps.scrape.output}
      Enrichment data: {steps.enrich.output}
      Consider: company size, funding stage, tech stack sophistication, growth signals.
      Return JSON with score and reasoning.
    model: haiku
    max_turns: 5
    output_schema:
      type: object
      properties:
        score: { type: integer }
        reasoning: { type: string }
        signals:
          type: array
          items: { type: string }

on_complete:
  webhook: ${CALLBACK_URL}
  storage_path: enrichments/{run_id}/results.json

on_failure:
  dead_letter: true
  webhook: ${FAILURE_WEBHOOK_URL}

schedule: null