Docula is a deterministic Bayesian NLP workstation built for engineers, DevOps, and legal counsel. Ingest massive logs, CSV dumps, multi-sheet spreadsheets, and legal archives into an ultra-fast CSR knowledge graph with line-exact citation proof.
Trace indirect causal chains ($A \rightarrow B \rightarrow C$) across microservice logs and journal dumps to find the root cause of crashes.
Process high-scale CSV exports, tabular test matrices, and raw technical documentation with zero cloud token billing.
Index thousands of PDFs, Word files, and legacy documents with 100% mathematically reproducible citation grounding.
Ingest multi-sheet `.xlsx` and `.xls` workbooks into structured tabular CSV streams to extract contrastive market data.
Why systems architects, forensic engineers, and legal auditors choose Docula over stochastic cloud LLMs.
Never take a summary at face value. Docula couples its Bayesian evidence scoring directly with an interactive document viewer.
Clicking any ranked citation immediately opens the full original file, displays line gutters, and jumps straight to the target line with synchronized bi-directional search (Next / Prev).
Organize your investigations into independent project workspaces. Switching projects instantly resets the view and loads a fresh, unbranded suggestion pool mined directly from that project's graph.
Every submitted inquiry, synthesized conclusion, and score-ranked citation set is committed to a local JSON history database for auditability and rapid recall.
Engineered from first principles without bloated dependencies or brittle cloud APIs.
Ingests system journals, syslog streams, and application traces. Finds latent bridge nodes connecting sporadic fault codes to root configuration changes.
Identifies indirect causal links ($A \rightarrow B \rightarrow C$) across disconnected files where concepts never appear in the same sentence, maximizing joint transition path weights.
Evaluates set differences across statistical neighborhoods ($A \text{ vs. } B$) to clearly articulate how competing parameters, clauses, or data sets differ in practice.
Processes Word (`.docx`) through direct XML run parsing and translates complex multi-sheet Excel (`.xlsx`, `.xls`) workbooks into normalized flat CSV with sheet annotations.
Stores the entire Bayesian transition graph in flat contiguous memory arrays without pointer indirection, unlocking multi-threaded CPU parallel processing at hardware bandwidth limits.
Engineered for mission-critical legal discovery, financial forensics, defense analysis, and private IP research. No data packets ever leave your host system.
Continuous engine upgrades, renewable license key protection via Gumroad, and multi-platform coverage.