Configuration

Vizzor reads config from ~/.vizzor/config.yaml. Environment variables override file values, so the same image can run in dev, staging, and production with no file edits.

Sample config

anthropicApiKey: <your-key>
etherscanApiKey: <your-key>
defaultChain: ethereum
 
rpc:
  ethereum: https://eth-mainnet.g.alchemy.com/v2/<your-key>
  polygon: https://polygon-mainnet.g.alchemy.com/v2/<your-key>
 
ai:
  provider: anthropic        # anthropic | openai | gemini | ollama
  model: claude-sonnet-4-20250514
  maxTokens: 4096
  ollamaHost: http://localhost:11434
 
output:
  format: table
  color: true
  verbose: false
 
cacheTtl:
  tokenInfo: 3600
  marketData: 300
  walletData: 600
  contractCode: 86400
 
polymarket:
  enabled: true
  defaultSize: 10          # USD per trade
  orderType: limit

AI provider comparison

API keys

No API key at all? Use Ollama with a local model — fully offline predictions (lower quality but free).

Runtime knobs (no YAML edit required)

Every operator-tunable setting is reachable from chat. The user rule is explicit: "edit YAML" is treated as "hardcoded". Symbols, slots, thresholds, mode, risk profile, allowlist — all settable at runtime from Telegram (admins) or the CLI.

1# Scheduler — slot-level knobs2/scheduler on3/scheduler quality-gate scalp-30m on conf=65 floor=partial4/scheduler vol-adapter scalp-30m on5/scheduler auto-disable scalp-30m on floor=0.35 window=30 breaches=36/scheduler horizons scalp-30m 5m 15m 30m7/scheduler symbols scalp-30m BTC ETH SOL8 9# Engine mode10/mode balanced               # balanced | precision | sniper | cosmos11 12# Per-user13/tz Europe/Madrid14/quiet 22:00-08:0015/notify_threshold tracked16 17# Runtime allowlist (admin only)18/allow 123456789019/deny 987654321020/access

These changes:

• Take effect immediately (no restart)
• Persist via the vizzor_state SQLite volume
• Survive bot restarts
• Are mirrored back on /health and /scheduler for inspection

Cache TTLs

Cache TTLs apply to the SQLite cache that fronts every external API. The defaults are deliberately conservative — tune them per workload:

Setting any TTL to 0 disables the cache for that bucket.