FilesExpand file tree

main

/

config.example.yaml

Copy path

Blame

More file actions

Blame

More file actions

Latest commit

 

History

History

History

185 lines (171 loc) · 7.32 KB

main

/

config.example.yaml

Copy path

Top

File metadata and controls

• Code
• Blame

185 lines (171 loc) · 7.32 KB

Raw

Copy raw file

Download raw file

Open symbols panel

Edit and raw actions

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

97

98

99

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

# ============================================================================

# Intel Sweep — Competitive Intelligence Scanner

# Configuration Reference

# ============================================================================

# Copy this file to config.yaml and customize for your use case.

# All API keys should be set via environment variables, never in this file.

# ============================================================================

# --- Identity ---

project:

name: "my-intel-sweep"

description: "Competitive landscape monitoring for [your domain]"

# --- Topics to Monitor ---

# Each topic defines a monitoring track with its own keywords,

# relevance criteria, and priority level.

topics:

- id: competitors

name: "Competitor Moves"

description: "Track funding rounds, product launches, acquisitions, and leadership changes"

keywords:

- "Harmonic.ai"

- "Sourcescrub"

- "SignalFire Beacon"

- "PitchBook automation"

relevance_prompt: |

Score this article for relevance to competitive intelligence in the

PE/VC due diligence tooling space. High relevance: direct competitor

funding, product launches, pivots, acquisitions, or leadership changes.

Medium: adjacent market moves. Low: tangential mentions.

priority: high

schedule: "0 7 * * *" # daily at 7am UTC

- id: market_signals

name: "Market Signals"

description: "Industry trends, regulatory changes, analyst reports"

keywords:

- "AI due diligence"

- "entity verification"

- "APAC startup ecosystem"

- "PE VC technology"

relevance_prompt: |

Score this article for strategic relevance to AI-powered entity

verification and startup intelligence platforms operating in

APAC markets. High: regulatory changes, market sizing, analyst

reports. Medium: adjacent technology trends. Low: general AI news.

priority: medium

schedule: "0 7 * * 1,4" # Mon and Thu at 7am UTC

- id: tech_patterns

name: "Architecture Patterns"

description: "Agentic AI, MCP protocols, verification infrastructure"

keywords:

- "agentic architecture"

- "MCP protocol AI"

- "LLM verification"

- "hallucination detection"

relevance_prompt: |

Score this article for technical relevance to building AI agent

systems with verification layers, confidence scoring, and

source-lineage tracking. High: novel architectures, benchmark

results, production case studies. Medium: tutorials, opinion

pieces. Low: hype articles with no technical substance.

priority: low

schedule: "0 7 * * 5" # Fridays at 7am UTC

# --- Search Configuration ---

search:

provider: "google" # google | serper | brave | tavily

max_results_per_keyword: 5 # results per keyword per run

lookback_hours: 168 # only surface content from last N hours

deduplicate: true # skip URLs already seen

# provider-specific settings loaded from env:

# SEARCH_API_KEY

# --- Scoring ---

scoring:

provider: "gemini" # gemini | openai | anthropic | ollama

model: "gemini-2.5-flash" # model string for scoring LLM

temperature: 0.1 # low temp for consistent scoring

threshold:

surface: 0.6 # minimum score to surface to user

auto_approve: 0.9 # above this, skip manual approval

# provider-specific settings loaded from env:

# SCORING_API_KEY

# OLLAMA_BASE_URL (if provider: ollama)

# --- Notifications ---

notifications:

- channel: slack

webhook_env: "SLACK_WEBHOOK_URL" # env var name, not the URL itself

format: "compact" # compact | detailed

include_score: true

include_summary: true

approval_enabled: true # add approve/discard reactions

# Optional: filter which topics go to this channel

topics: ["competitors", "market_signals"]

- channel: telegram

bot_token_env: "TELEGRAM_BOT_TOKEN"

chat_id_env: "TELEGRAM_CHAT_ID"

format: "detailed"

topics: ["tech_patterns"]

# --- Approval & Storage ---

approval:

method: "reaction" # reaction | webhook | cli

approved_label: "thumbsup" # emoji reaction = approved

discarded_label: "thumbsdown" # emoji reaction = discarded

auto_expire_hours: 72 # un-reviewed items expire after N hours

storage:

backend: "firestore" # firestore | local_json | sqlite

# Firestore settings

collection: "intel_sweep_items"

# Local settings

local_path: "./data/items.json"

# Approved items accumulate here for downstream consumption

approved_output:

format: "jsonl" # jsonl | markdown | both

path: "./data/approved.jsonl"

# Optional: also write to a Google Doc for Claude context

gdoc_id: "" # leave empty to disable

# --- Authenticity Detection ---

# Three-layer AI content authenticity pipeline (see src/authenticity.py).

# Layer 1: Source reputation (YAML-driven domain tiers, zero cost)

# Layer 2: Content heuristics (7 deterministic Python checks, zero cost)

# Layer 3: LLM authenticity detection (optional, 1 API call per article)

authenticity:

min_score: 0.4 # Block items below this composite score

auto_approve_min_score: 0.8 # Require this score for auto-approval (alongside relevance)

use_llm_layer: false # Enable Layer 3 LLM check (uses scoring API key, costs money)

source_reputation_path: "config/source_reputation.yaml" # Domain trust DB

# Optional: override composite scoring weights (values must sum to 1.0 per group)

# weights:

# no_llm: # used when use_llm_layer: false (defaults: source 0.45, heuristic 0.55)

# source: 0.45

# heuristic: 0.55

# with_llm: # used when use_llm_layer: true (defaults: source 0.30, heuristic 0.35, llm 0.35)

# source: 0.30

# heuristic: 0.35

# llm: 0.35

# --- Gamut Intelligence Integration (Optional) ---

# If you have Gamut API credentials, discovered entities can be

# automatically verified against APAC registries with confidence scoring.

gamut:

enabled: false

api_base: "https://api.gamutagent.ai/v1"

api_key_env: "GAMUT_API_KEY"

auto_verify_entities: true # run gamut_verify_entity on discovered companies

attach_confidence_score: true # include CVI in surfaced results

# --- Secrets ---

# How API keys and tokens are resolved. The config file contains only

# the NAMES of secrets, never values.

secrets:

backend: "env" # env | dotenv | gcp | aws | azure

# dotenv settings:

# dotenv_path: ".env"

# GCP settings (also uses GCP_PROJECT_ID env var):

# project_id: "my-project"

# secret_prefix: "intel-sweep-"

# AWS settings (also uses AWS_REGION env var):

# region: "us-east-1"

# secret_prefix: "intel-sweep/"

# Azure settings (also uses AZURE_VAULT_URL env var):

# vault_url: "https://my-vault.vault.azure.net"

# --- Security ---

security:

# Bind to localhost only — never expose to 0.0.0.0

bind_address: "127.0.0.1"

# Rate limiting

max_searches_per_hour: 50

max_scoring_calls_per_hour: 200

# Content safety

block_domains: # never fetch content from these

- "malware-site.example"

allow_domains: [] # if non-empty, ONLY fetch from these

# Audit

log_all_searches: true

log_scoring_decisions: true

audit_log_path: "./logs/audit.jsonl"