Galadril ⛲️

Documentation | GitHub

"Things that were, and things that are, and some things that have not yet come to pass."

Galadril is an advanced data integration and analytical intelligence platform designed to provide a "Mirror" of complex systems. Galadril focuses on elucidation, foresight, and transparency.

Caution

This project is still in its early stages.

Development

Enter the shell to load the environment:

nix develop github:RealHinome/Galadril?dir=infrastructure/nix

Targeted architecture

flowchart TD
    classDef source fill:#e3f2fd,stroke:#1565c0,stroke-width:2px,color:#0d47a1
    classDef ingest fill:#ffecb3,stroke:#ff6f00,stroke-width:2px,color:#3e2723
    classDef stream fill:#d1c4e9,stroke:#512da8,stroke-width:2px,color:#311b92
    classDef ml fill:#f8bbd0,stroke:#c2185b,stroke-width:2px,color:#880e4f
    classDef pg fill:#336791,stroke:#000,stroke-width:2px,color:#fff
    classDef app fill:#c8e6c9,stroke:#2e7d32,stroke-width:2px,color:#1b5e20
    classDef bus fill:#212121,stroke:#fff,stroke-width:2px,color:#fff,stroke-dasharray: 5 5

    subgraph Sources ["Multi-Modal Data Sources"]
        direction TB
        S1[("Sensors / IoT / SIGINT")]:::source
        S2[("Financial / ERP Flows")]:::source
        S3[("Unstructured (OSINT/Docs)")]:::source
        S4[("3rd Party APIs")]:::source
    end

    subgraph Ingestion ["Ingestor"]
        Connectors["Smart Connectors"]:::ingest
        Raw_Bus[("Raw Event Bus (Kafka)")]:::bus
    end

    subgraph Processing ["The Vision"]
        Stream_Engine["Stream Processor"]:::stream

        subgraph Compute ["Compute Services"]
            Entity_Res["Entity Resolution"]:::ml
            Ontology_Map["Ontology Mapper"]:::stream
            ML_Inf["ML Inference"]:::ml
        end

        Feature_Store["Feature Store (Online)"]:::pg
    end

    subgraph Knowledge ["The Synapse"]
        Intel_Bus[("Curated Intel Bus (Kafka)")]:::bus

        subgraph PG_Engine ["PostgreSQL"]
            direction TB
            KG[("Apache AGE")]:::pg
            VecDB[("pgvectorscale")]:::pg
            Relational[("Relational")]:::pg
            Timescale[("TimescaleDB")]:::pg
        end

        ObjStore[("Object Store")]:::pg
    end

    subgraph Consumption ["Galadril Studio"]
        Gateway["Unified Ontology API"]:::app
        Studio["Investigation Graphs"]:::app
        Alerts["Operational Alerting"]:::app
    end

    S1 & S2 & S3 & S4 --> Connectors
    Connectors --> Raw_Bus
    Connectors -->|Direct Backup| ObjStore

    Raw_Bus --> Stream_Engine

    Stream_Engine <--> Ontology_Map
    Stream_Engine <--> Feature_Store

    Feature_Store -.-> |"Get Features"| ML_Inf
    ML_Inf --> Stream_Engine

    Entity_Res <--> |"Lookup / Match"| Relational
    Stream_Engine <--> Entity_Res

    Stream_Engine --> Intel_Bus
    Intel_Bus --> PG_Engine

    PG_Engine & ObjStore --> Gateway
    Gateway <--> Studio
    Gateway --> Alerts

Loading

ESKG-enhanced GraphRAG

Galadril implements a reasoning framework based on the Event-State Knowledge Graph (ESKG), as described in Zang et al. (2026).

Its core transition pattern is:

$$ S_t \xrightarrow{Triggers} E_t \xrightarrow{LeadsTo} S_{t+1} $$

A state enables an event, the event produces a new state, and the graph grows with this history. This makes state changes traceable rather than simply overwriting the previous value.

The base ESKG defines six relations:

Relation	Meaning
Triggers ($S \rightarrow E$)	A state enables or activates an event.
LeadsTo ($E \rightarrow S$)	An event produces a resulting state.
Evolution ($S \rightarrow S$)	A state directly evolves into another state.
Contain ($E \rightarrow E$)	An event contains or decomposes into other events.
Occur ($E \rightarrow P$)	An event occurs at or is associated with a physical entity.
Influence ($E \rightarrow P$)	An event affects a physical entity or one of its properties.

Galadril extends this model with Latent Identity ESKG. It therefore keeps evidence and identity hypotheses separate from authoritative graph facts until a resolution decision is made.

Its additional relations describe that resolution process:

Relation	Meaning
hasInference	Links an observation to its inference result.
consideredCandidate	Records identities considered during inference.
hasDecision	Links an inference to the resulting decision.
selectedTarget	Records the identity selected by that decision.
promotedAs	Resolves a latent identity into an authoritative entity.
mergedWith	Versionably merges two latent identity hypotheses.
supersedes / revokes	Replaces or invalidates an earlier interpretation without deleting history.

License

This project is licensed under the terms of the BSD 3-Clause License. See the LICENSE file for the full license text.