Reasoning you can inspect.
CHIAKOR turns macro-financial cause and effect into an explicit knowledge graph — every recommendation is a traceable chain of signed, weighted, provenance-tagged evidence.
- Point-in-time correct
- Evidence on every call
- Vendor-neutral engine
POST /v1/reasoninflation ↑real_rates ↑usd_liquidity ↑credit_spread ↑From raw signals to a decision you can defend
A single, transparent pipeline. Follow any recommendation back to the evidence that produced it.
One reasoning engine powers every product
The same explainable signal branches into every downstream experience — trading styles, portfolio intelligence, delivery and integrations. Live today or clearly on the Phase 6 roadmap.
Trading styles
- Scalping signalsP6
- Swing tradingP6
- Position tradingP6
- Long-term investingP6
Portfolio & risk
- Portfolio optimizationLive
- Risk analysisP6
Delivery
- Trading signalsLive
- NotificationsLive
Integrations
- Developer APILive
- MetaTrader bridgeP6
- Broker executionP6
Built for reasoning, not black boxes
Every capability below is live in the engine today.
Explainable by design
The explanation is the reasoning — not a summary bolted on afterwards.
Point-in-time correct
A temporal graph with no look-ahead. Decisions use only what was known then.
Evidence & provenance
Every influence path is signed, weighted and traceable to its source.
Backtesting
Replay scenarios over history and calibrate against real outcomes.
Vendor-neutral
A frozen engine behind clean APIs, with pluggable language models.
Multi-tenant & RBAC
JWT identity, role-based permissions and per-tenant isolation.
Everything CHIAKOR does, connected
One reasoning core powers every surface — from explainable signals to portfolio intelligence. Each capability traces back to the same evidence you just watched move through the graph.
Cause and effect, made explicit
CHIAKOR models the market as a graph of signed, weighted relationships between factors and assets — so reasoning is inspection, not guesswork.
- Canonical entities and typed, signed relations.
- Temporal edges — the graph knows what was true, and when.
- Every edge carries a strength and a provenance record.
Trustworthy by construction
The properties that make the reasoning safe to rely on.
Identity & RBAC
HS256 JWT, role-based permissions, per-tenant isolation.
No look-ahead
Point-in-time correctness is enforced across the graph.
Full provenance
Every contribution path records where its evidence came from.
Observability
Health, readiness, metrics and rate limiting are built in.
A clean REST API over the reasoning engine
Authenticate once, then reason, explain, gather evidence and backtest over stable endpoints.
/v1/reason/v1/explain/v1/evidence/v1/backtest/v1/capabilities/v1/health# 1) authenticate
curl -X POST $API/v1/auth/login \
-d '{"user":"analyst","password":"…"}'
# → { "token": "…" }
# 2) reason under a scenario
curl -X POST $API/v1/reason \
-H "Authorization: Bearer $TOKEN" \
-d '{"scenario":{"inflation":"rising",
"real_rates":"rising"}}'Reasoning you can see, not a number you can't
Most models output a score and hide the why. CHIAKOR shows the entire chain.
Typical black-box model
- A single opaque score
- No traceable evidence
- Look-ahead bias is easy to hide
- Hard to audit or defend
CHIAKOR
- Signed, weighted evidence paths
- Full provenance on every number
- Point-in-time correct by construction
- Auditable and explainable end to end
What teams would say
How CHIAKOR fits real research and portfolio workflows.
“Finally a model where I can follow the reasoning instead of trusting a black box.”
“The evidence paths make it trivial to defend a call in a review.”
“Point-in-time correctness is the feature I didn't know I needed.”
Illustrative personas — CHIAKOR is a demonstrator and these are not real customer quotes.
Frequently asked questions
The essentials, answered honestly.
No. CHIAKOR is an explainable-reasoning demonstrator. Confidence is uncalibrated and must never size real positions.
Every recommendation decomposes into signed, weighted influence paths with provenance — the explanation is the reasoning itself, not a post-hoc narrative.
The reasoning engine is model-agnostic. Language models are pluggable behind a vendor-neutral seam and never replace the deterministic core.
Yes. The temporal knowledge graph enforces no look-ahead — a decision only uses information available at that moment.
CHIAKOR is the product experience; FinKG is the frozen reasoning engine underneath it, exposed through the platform APIs.