Resources · Secret Sauce
How xChat gives higher-quality answers
xChat performs best when it combines your workspace context (portfolio + account settings), scoring factors, and mixed retrieval (collections + web/x-search + live market checks) in one tool loop. This page explains the mechanics in plain language.
1) Workspace context: portfolio, outlook, risk, scoring factors
xChat is not answering in a vacuum. It uses your workspace state so responses match your book and constraints, not generic internet commentary.
- Portfolio + account context: default portfolio/account establish where positions and cash risk sit.
- Outlook + risk profile: strategy direction and aggressiveness are filtered to match your desk stance.
- SE Scoring Factors: scoring weights prioritize what your workspace values (for example liquidity, vol profile, income-fit).
2) Search blend: collections + xChat + web + live x-search
xChat combines internal and external sources to reduce blind spots and stale reasoning.
Collections search (your docs)
Pulls from configured workspace collections first, so policy notes, desk memos, and internal standards are grounded in your own source material.
Web/x-search (external context)
Adds current public context when needed (news flow, market narrative, recent developments), instead of relying only on local docs.
Live market checks
For symbols/chains, xChat can route to live quote or market tools so strike/expiry reasoning uses fresher numbers.
Single answer synthesis
The model merges these signals into one recommendation path, with rationale aligned to your workspace constraints.
3) Tool loop (brief)
xChat runs an iterative loop: think -> call tools -> absorb results -> continue until it can answer with enough confidence.
- User prompt enters with persona + workspace context.
- Model decides which tools to call (collections, web/x-search, market/tooling).
- Tool outputs feed back into the same response run.
- Loop repeats until completion and then returns a final response.
4) xStrategyBuilder + engine concepts
xStrategyBuilder and the strategy engine apply the same context discipline in a more explicit pipeline.
- User context stage: portfolio/account + risk/outlook + scoring factors shape which strategies are valid.
- Symbol/chain stage: live contract rows are evaluated for feasibility and quality.
- Fit + ranking stage: candidates are scored, ranked, and packaged with risk/reward rationale.
Explore the live flow in xStrategyBuilder, then pressure-test assumptions in xChat.
5) How to get the best responses
- State your objective first (income, hedge, directional, assignment tolerance).
- Give timeframe + risk budget + position constraints in the prompt.
- Ask for alternatives: base case, conservative case, and risk-off adjustment.
- Request explicit assumptions and failure conditions before execution decisions.