Seer Shared

What Seer Is

• Claims — atomic evidence statements extracted from sources (the raw events/facts)
• Threads — tradable angles / sub-theses built from clusters of claims
• Topics — narrative domains that group related threads (e.g., "AI Infrastructure", "Energy Transition")
• Global snapshot — overall market regime derived from top drivers across all topics

seer-q

The Web UI

view_url

• Topic sphere — click any topic node to zoom into its threads, then drill down to individual claims
• Driver graph — causal links between market drivers as a force-directed network you can rotate and explore
• Verdict rail — real-time regime summary with conviction level and key uncertainties

view_url

Response Format

• Success (stdout):

  { "ok": true, "data": <payload>, "meta": { "view_url": "...", "count": N } }

• Errors (stderr):

  { "ok": false, "error": { "code": "...", "message": "..." } }

.data

view_url

.meta.view_url

.meta

.data

view_url

.data

--raw

Global Flags

--format

json

json

pretty

--raw

--api-url

Config & Diagnostics

seer-q version                  # CLI version, Go version, OS/arch
seer-q doctor                   # Verify API connectivity and config
seer-q config get <key>         # Get config value
seer-q config set <key> <value> # Set config value
seer-q config list              # Show active configuration
seer-q config path              # Show config file path
seer-q schema                   # List all command contracts
seer-q schema <command>         # Describe a command's input/output contract
seer-q health                   # API health check

Common Rules

1. Use

   seer-q search

   first whenever the user mentions a specific entity, asset, or theme
2. Always include

   view_url

   as clickable markdown links. The page-level URL is in

   .meta.view_url

   . Per-entity URLs (topics, threads) are on each object inside

   .data

   . Format as

   [descriptive text](url)

   — never paste raw URLs. Example:

   [Explore this topic on the interactive map](https://www.midaz.xyz/market?topic=abc123)

   .
3. Synthesize data into natural language — don't dump raw JSON
4. For multi-entity questions, make multiple calls to build a complete picture
5. When claims are asked about, note their

   claim_mode

   ,

   thread_role

   (support/contradiction), and

   event_date