Oblivion / laboratory surfaceANGRA / integration layer — developer releasev0.1
OblivionAngra
Connect the models you already run. ANGRA links compatible AI models running through LM Studio and Bionic so specialized model instances can exchange tasks, results and context in one coordinated workflow.
One model becomes a worker. A network of models becomes a workforce.
ManifestoRead the figures in order
A model generates.
ANGRA coordinates.
A model generates. ANGRA coordinates. A network collaborates. A workflow becomes a system.
AI becomes more useful when specialized models can cooperate instead of working as isolated interfaces. The emphasis is coordination, specialization, interoperability and networking, with humans keeping direction and final control.
One model is a worker.
Many models are a workforce.
System / 01ANGRA
Isolated model.
Coordinated work.
An isolated model turns a prompt into a response. ANGRA is an Oblivion-built layer that connects compatible models from your LM Studio and Bionic environments so they can divide a task. ANGRA contains no models and does not replace either environment.
LayerIntegration
TargetsLM Studio · Bionic
ReleaseDeveloper
ModelsYours
NetworkDesigned
Version1.0.0
Module 01 / PipelinesIllustrative
- Isolated model
- Prompt
- Inference
- Response
- ANGRA network
- Task
- Delegate
- Route
- Collaborate
- Exchange
- Verify
- Converge
- Result
Module 02 / Task graph — one objective, many workflowsSimulated
Tap a node to branch it.
PassWaitingRetryReview
Module 03 / Coordination loop — Conceptual
Observe
- What happens
- Computation
- Tools
- Next
Module 04 / Architecture — hover or tap a nodeConceptual
Select a node
Each node is a stage of working state. Upstream stages are outlined in amber, downstream in orange.
Module 05 / Connectors — configured, scoped, auditable
Connectors are designed around operator-controlled connectivity: only compatible runtimes and supported endpoints that you configure. Nothing in this page is connected to a live runtime.
Operator remains in control
Coordinating
Workforce / NetworkConceptual — no live network in this page
From isolated models
to a connected workforce.
A single local model is useful. Several specialized models on one machine can divide work. Several model runtimes across configured nodes can form a larger workflow. ANGRA is the coordination layer between those workers: task, route, handoff, result, verification.
“AI employees” is an organizational metaphor for specialized model instances with different responsibilities. LM Studio and Bionic supply the environments and models; ANGRA connects compatible runtimes and supported endpoints. Oblivion is not affiliated with or endorsed by either.
Multi-model workflowConceptual
- Objective
- ANGRA
- Task decomposition
- Specialized model selection
- Parallel / sequential execution
- Model-to-model handoff
- Result merge
- Verification
- Final output
Security principlesDesigned for ≠ implemented
Designed forNot claimed
- Authenticated nodes and explicit peer trust
- Encrypted transport when configured
- Scoped permissions and task authorization
- Least privilege, rate limiting, message validation
- Replay protection where implemented
- Fail-closed behavior and auditability
Implemented in 1.0.0From the scripts
- Local operation; no server started and no network traffic sent by either script
- Only models you authorize or have already loaded
- Timeouts, cooldowns and output limits
- Local logging
Networked model collaboration is not part of 1.0.0.
Optional deployment controlsOperator-owned
- Operator-controlled connectivity
- Network segmentation and firewall rules
- Secure transport chosen by the deployer
No claim of complete security is made.
Download / ANGRATwo packages · served from this site
Get ANGRA.
Direct static downloads; no redirect, no simulated progress. The SHA-256 and size shown are computed from the files in /downloads/. ANGRA needs a separately installed LM Studio or Bionic.
Demos / WorkflowsAll sequences illustrative
One objective.
Many operations.
Work treated as a routing problem: sub-tasks, workers and handoffs become structure before they become a result. Sequences below are simulated.
Module 06 / RecursionProcedural halftone
The team builds the system.
The system builds with the team.
Oblivion engineers use AI workers while developing software and games. ANGRA is the coordination layer between those workers; the humans own architecture, direction, review and final control.
Humans + ANGRA workers + Software + Games → Oblivion products
Research / Evaluation labNo numbers are invented here
Evidence
before claims.
Multi-model coordinationLocal AI orchestrationModel interoperabilityAgent collaborationDistributed AI workflowsAI-assisted software engineeringHuman + AI development
Data lives in EVAL in the page script. A field left null renders as NOT PUBLISHED.
BulletinsEdit NEWS / ARTICLES to publish
TeamOrganizational model — not live runtime data
Oblivion
team
Oblivion is an independent software, AI systems and game development organization. Its current organizational model is about 104 members: 4 human programmers and 100 AI-agent workers. The 100 are roles or model/agent instances coordinated through ANGRA, not people, and this page does not claim they all run at once.
Founder / leadBardiya Shokri
ProgrammerSahand Farzaneh
ProgrammerBarsam Alizadeh
ProgrammerDanial Dadashi
ANGRA workforce100 AI workers
ANGRA organizational model — specialized AI workers
ResearchEngineeringCodeTestingDataDocumentationAnalysisSecurityGame devPlanningReviewOperations
FAQFactual answers only
ContactDeveloper release · Operator-controlled
Request
information
Contact Oblivion
Address is a placeholder until configured. ANGRA 1.0.0 is a developer release.