Playground Pricing Technology Docs Dashboard Contact Start Free Trial
Financial SEC API

Deterministic
SEC Filing
Intelligence.

Zero-hallucination, provenance-backed answers for financial analysis. When the lattice cannot certify a result, it returns an honest silence. We don't guess.

0.00% Hallucination Rate
100% Determinism
0 LLMs Involved
92% FinanceBench Recall

Live API

Financial SEC — SEC Filing QA ● Checking…

Ask anything about a 10-K, 10-Q, or 8-K. Answers are grounded to a specific filing, page, and XBRL concept — or the API returns honest silence. Try it live.

Checking… financesec.api.caleralabs.com
ACTIVE RUN: BAKE-12

Lattice Training Dashboard

8 / 15 Epoch
0.0124 Loss
92.1% FB Recall
10.2M Weights

The FinSec Volumetric Lattice Network is online and fully synchronized with the SEC EDGAR filing database. Traversals, mathematical decompositions, and cryptographic grounding provenance are fully live.

Query Taxonomy — What the Lattice Handles

The VLN classifies every query into one of four categories. Click any example to load it into the console. Every category is a feature, not a limitation — even honest silence is a guarantee no other system offers.

LIVE
Direct Recall

Single-fact answers decoded from frozen SmartWeights with full XBRL provenance chain.

LATTICE_RECALL

SDR encoding → A₄ coordinate walk → Hopfield settling at T=0 → exact basin match. Deterministic: same query, same answer, every time.

LIVE
Analytical Decomposition

Multi-fact ratios and growth rates computed on-the-fly by decomposing into atomic recalls.

Gross Profit ÷ Revenue = Margin
Both facts recalled independently, then computed deterministically.
ANALYTICAL_DECOMPOSER

Decomposes query → recalls component facts → performs arithmetic via domain-layer physics. Each component carries its own provenance chain.

LIVE
Tentative Analysis

Qualitative patterns and judgment calls with explicit caveats — the lattice found evidence but flags uncertainty.

CONFIDENCE_GATE v2.9

When confidence falls between 40–75%, the lattice returns the answer with a TENTATIVE_MATCH flag so you know the evidence is not fully crystallized.

BY DESIGN
Honest Silence

Queries the lattice has no crystallized evidence for — and therefore will never fabricate an answer to.

SAFE_REFUSAL

The lattice physically resonates only with evidence it has crystallized. If the resonance pattern is absent, no output is produced. This is a feature, not a failure.

What Can I Ask?

The Financial SEC API answers questions about SEC filings for 10,000+ public companies in the EDGAR universe. Here's the full range of supported query types:

📊
Financial Metrics

Revenue, net income, gross profit, total assets, operating income, EPS, and all standard XBRL-tagged financial statement items.

Example: "What was Microsoft's net income in FY2022?"
📐
Financial Ratios

Gross margin, operating margin, quick ratio, current ratio, debt-to-equity, ROE, ROA, asset turnover, and DPO — computed from source facts.

Example: "What is AMD's FY2022 quick ratio?"
📈
Growth Rates

Year-over-year revenue growth, net income growth, capex growth, and operating income changes between any two ingested fiscal years.

Example: "What is Amazon's year-over-year change in revenue from FY2016 to FY2017?"
📋
Filing Metadata

Key agendas from 8-K filings, geographic operating regions, registered debt securities, and other structured filing metadata.

Example: "What was the key agenda of Amcor's 8K filing dated 1st July 2022?"
🔍
Qualitative Analysis

Capital intensity assessments, dividend trends, operating margin drivers, and domain-relevant analytical judgments grounded to filing evidence.

Example: "Is 3M a capital-intensive business based on FY2022 data?"
🛡️
Provenance & Audit

Every answer includes the source filing, XBRL concept, document period, page number, and a SHA-256 audit hash for reproducibility.

Example: Click any answer's "Provenance Chain" to see its full audit trail.

Example Queries — FinanceBench Gold Answers

Company
Metric
Period
Value
Confidence

Click any row to load that query into the playground above — company selector and question are pre-filled automatically.

Lattice Statistics

0
Hallucinations
Zero wrong answers on FinanceBench
100%
Precision
Every answer given was verified correct
92%
Recall
138 of 150 FinanceBench questions answered
100%
Coverage
Full coverage of indexed entities
v0.1.7
Version
Platform version
d222effd29c2…
Battery SHA
Reproducibility manifest hash

Looking for other specialized APIs? Visit the CaleraLabs Product Hub →

Pricing

Simple, Transparent API Pricing

Monthly Annual 20% Off
Feature
Pilot
Free
14-day trial
Hacker
$99/mo
Plus applicable taxes
Indie devs
Startup
$299/mo
Plus applicable taxes
Small teams
Most Popular
Starter
$2,000/mo
Plus applicable taxes
PE / VC firms
Professional
$5,000/mo
Plus applicable taxes
Hedge funds
Enterprise
$10,000/mo
Plus applicable taxes
Institutions
Usage
Queries / day 1,000 2,000 8,000 250,000 Unlimited
Company coverage Full EDGAR Full EDGAR Full EDGAR Full EDGAR Full + Custom
Filing types 10-K 10-K, 10-Q 10-K, 10-Q 10-K, 10-Q, 8-K All SEC filings
Concurrent connections 1 2 5 25 Unlimited
Core Guarantees
Zero hallucination
Full provenance chain
Deterministic recall
API Features
Query endpoint
Provenance API
Trend analysis API
Batch query API
Custom XBRL ingestion
Audit & Compliance
Audit log retention 14 days 30 days 90 days 365 days Unlimited
SHA-256 audit hash
Hardware attestation
Support & SLA
Support channel Email Email Email + Slack Dedicated Dedicated + Phone
Uptime SLA 99.5% 99.5% 99.95% 99.99%
Start Free Trial Get Started Get Started Contact Sales Contact Sales

All tiers include our zero hallucination guarantee — same engine, same architecture. No LLM in the answer path. Learn how →

The Substrate

Not a Wrapper Around an LLM

Every CaleraLabs API runs on the Volumetric Lattice Network — a locally-hosted associative memory substrate that learns via Hebbian reinforcement over a geometric lattice, not gradient descent over a transformer.

  • Provenance by construction: Every answer traces back to the source document and evidence span that produced it.
  • Honest refusal: If the lattice has no crystallized evidence for a query, it returns a safe refusal — never a fabricated guess.
  • Deterministic recall: The same query against the same trained model returns the identical answer, every time.
  • Analytical decomposition: Multi-fact queries (margins, ratios, growth rates) are decomposed into atomic recalls, then computed deterministically — each component carries its own provenance chain.
  • No data leaves the lattice: Training data and model weights are never sent to a third-party model provider.
NLU Decomposition
Natural language → intent, entity, and temporal atoms
A₄ Coordinate Walk
Geometric traversal through lattice hyperspace
Hopfield Settling
Energy minimization to nearest memory basin
Verified Result
Provenance-backed answer with audit hash

Getting Started

Pilot Access

Every product offers a free pilot tier with a time-limited license key. Sign in to your dashboard to view your API keys, usage analytics, and post-quantum security settings.

Sign In to Dashboard

Built by Calera Computing

CaleraLabs is the public API platform of Calera Computing, Inc. — the team behind the Volumetric Lattice Network.

🔬 Sandbox Simulator
Lattice Console POST /api/query
System Online

Welcome to the FinanceSec SEC Filing QA interface powered by the Volumetric Lattice Network. Ask queries about 10-Ks, 10-Qs, or 8-Ks.

Every response is strictly deterministic, verified with cryptographic grounding, and has 0.00% hallucination rate by design.