Echo voices and Baobab national models — Africa-first language intelligence.

Explore Baobab
Skip to content
Open consoleLog in

Lugemi status

Atlas engine

Multilingual reasoning engine catalogue and safety notes.

All status pagesRelated viewAll status pagesMachine JSON
product
Lugemi Atlas
note
Atlas. Interface scaffold for a large multilingual reasoning family. Capabilities map to existing Gateway/Reasoning Runtime/MLOps hubs. Does not ship trained Atlas weights, OpenAI replacement, or frontier-lab compute.
docs
/docs/ATLAS.md

capabilities

capabilities: 18

  • Reasoning

    id
    reasoning
    name
    Reasoning
    status
    shipped
    api
    GET /v1/reasoning-runtime/engine
    notes
    Handoff to Reasoning Runtime — not a custom Atlas reasoner.

  • Planning

    id
    planning
    name
    Planning
    status
    shipped
    api
    POST /v1/reasoning-runtime/plan
    notes
    Plan/reflect via Reasoning Runtime; no tool execution OS.

  • Coding

    id
    coding
    name
    Coding
    status
    deferred
    api
    null
    notes
    Code-specialist Atlas weights deferred.

  • Math

    id
    math
    name
    Math
    status
    deferred
    api
    null
    notes
    Math-specialist weights deferred.

  • Scientific Reasoning

    id
    scientific-reasoning
    name
    Scientific Reasoning
    status
    deferred
    api
    null
    notes
    Scientific specialist deferred.

  • Business Reasoning

    id
    business-reasoning
    name
    Business Reasoning
    status
    deferred
    api
    null
    notes
    Business specialist deferred.

  • Legal Reasoning

    id
    legal-reasoning
    name
    Legal Reasoning
    status
    deferred
    api
    null
    notes
    Legal specialist deferred — not legal advice product.

  • Medical Reasoning

    id
    medical-reasoning
    name
    Medical Reasoning
    status
    deferred
    api
    null
    notes
    Medical specialist deferred — not clinical product.

  • Financial Reasoning

    id
    financial-reasoning
    name
    Financial Reasoning
    status
    deferred
    api
    null
    notes
    Financial specialist deferred — not investment advice.

  • Multilingual

    id
    multilingual
    name
    Multilingual
    status
    shipped
    api
    POST /v1/chat/completions
    notes
    Vendor chat/Gateway today — not Atlas-trained multilingual weights.

  • Long Context

    id
    long-context
    name
    Long Context
    status
    shipped
    api
    POST /v1/context-runtime/assemble
    notes
    Context Runtime assemble/compress — not infinite-context OS.

  • Function Calling

    id
    function-calling
    name
    Function Calling
    status
    shipped
    api
    POST /v1/agent-runtime/runs
    notes
    Agent Runtime sandbox + Policy hard-gate — not open tools.

  • Tool Use

    id
    tool-use
    name
    Tool Use
    status
    shipped
    api
    POST /v1/agent-runtime/runs
    notes
    Same Agent Runtime sandbox path.

  • Training Pipeline

    id
    training-pipeline
    name
    Training Pipeline
    status
    shipped
    api
    POST /v1/model-training-platform/experiments
    notes
    Handoff to Model Training Platform /.

  • Inference

    id
    inference
    name
    Inference
    status
    shipped
    api
    POST /v1/chat/completions
    notes
    Gateway chat — not Atlas-hosted weights.

  • Evaluation

    id
    evaluation
    name
    Evaluation
    status
    shipped
    api
    POST /v1/model-evaluation-platform/runs
    notes
    Model Evaluation Platform — MMLU suite still deferred.

  • Benchmarking

    id
    benchmarking
    name
    Benchmarking
    status
    shipped
    api
    GET /v1/model-evaluation-platform/leaderboard
    notes
    Org-scoped sandbox ranks — not LMSYS.

  • Serving Platform

    id
    serving-platform
    name
    Serving Platform
    status
    shipped
    api
    GET /v1/model-serving/engine
    notes
    Model Serving hub — not Atlas cluster OS.

honesty

trainsCompetitiveFoundationWeights
false
shipsTrainedAtlasWeights
false
openAiReplacementOs
false
frontierLabOs
false
regeneratesPriorLayers
false
regeneratesReasoningRuntime
false
scaffoldOnly
true
extendsGateway
true
extendsMlopsTrack
true

architecture

style
nest_modular_monolith
ddd
bounded_atlas_scaffold
cqrs
true
hexagonalRewrite
false
repositoryPattern
handoffs_to_existing_modules
eventDriven
audit_and_jobs_only
solid
true
terraform
true
terraformPath
infra/DEPLOY.md
kubernetes
true
kubernetesPath
infra/AWS_EKS.md
primaryRegion
af-south-1
extendsGateway
true
extendsReasoningRuntime
true
extendsModelTrainingPlatform
true
extendsModelEvaluationPlatform
true
extendsFoundationModelCloud
true
regeneratesPriorLayers
false
trainsCompetitiveFoundationWeights
false
shipsTrainedAtlasWeights
false
openAiReplacementOs
false
frontierLabOs
false
customerFacingProduct
true
scaffoldOnly
true
note
Atlas as discoverable family scaffold. Real inference uses bought Gateway models until research charter + compute exist.

safety

noFakeTrainedWeights
true
scaffoldNotModel
true
note
Atlas is a family scaffold. Inference continues through Gateway vendors until trained weights exist.