AI Assistant — Claude
Not yet connected
🛡️ Public data only — no PHI permitted on this instance.
Anthropic Claude API
NOT CONFIGUREDBacks the AI-assist features on this platform — IC memo drafting, document Q&A, conversational portfolio queries. Every feature degrades gracefully when the key is absent.
Total calls
0
since install
Est. cost
$0.00
USD
Cached
0
prompts re-served
Models in use
0
seen in history
FEATURESwhat Claude powers on this platform
FEATURES THAT USE CLAUDE
| Feature | Endpoint | What it does |
|---|---|---|
| Partner review confirmation | /deal/<id>/partner-review | Runs a cached Claude second-pass over the PE verdict so the UI can show a concise confirm / watch-items summary. |
| IC Memo drafting | /api/deals/<id>/memo?llm=1 | Generates the memo narrative sections with fact-checking against packet dollar amounts and percentages. |
| Document Q&A | /api/deals/<id>/qa?q=… | Search indexed deal documents and return answers with confidence scores. |
| Multi-turn chat | POST /api/chat | Conversational interface with tool-calling — asks the platform for portfolio data and synthesizes answers. |
Every feature falls back to a non-LLM template when the key is not set, so the platform keeps working — just without synthesized prose.
MODELSClaude 4 lineup
AVAILABLE MODELS
| Model ID | Name | Notes |
|---|---|---|
| claude-opus-4-7 | Opus 4.7 | Highest-capability; deep reasoning. |
| claude-sonnet-4-6DEFAULT | Sonnet 4.6 | Balanced cost / quality. Default. |
| claude-haiku-4-5-20251001 | Haiku 4.5 | Fast / cheap. Good for bulk fact-checking. |
Override per-feature via the ``model`` kwarg on
LLMClient.generate() or set the ANTHROPIC_DEFAULT_MODEL env var.SETUPhow to connect your key
HOW TO CONNECT
- Get an API key from console.anthropic.com.
- Export it in the shell before launching the server:
export ANTHROPIC_API_KEY="sk-ant-…" .venv/bin/python seekingchartis.py --port 8090
- Reload this page. The status badge will flip to CONNECTED and the key fingerprint will show at the top of this panel.
The key is read from process env only — never stored on disk or logged. Cost tracking in
llm_calls records model + token counts, not prompts.Related
IntegrationsAutomation RulesAPI DocsAudit LogSystem Info