docs
Documentation
Slide decks for the bigger picture, and Key Topics for how each part of the platform actually works.
Slide decks
Presentation-style overviews — open any deck full-screen.
Agentic Apps: From Demo to Industrial Reality
The architectural patterns behind building production-grade agentic applications with Fred.
Fred: The Open Agentic Platform for the Enterprise
A high-level value-proposition deck for first-contact / innovation conversations.
Deploying Fred: A GitOps Deployment Pattern
The reference deployment design — a frozen infrastructure layer plus a GitOps-managed application layer.
Capabilities: One Switch For Everything An Agent Can Do
Fred's capability model for teams and architects: what ships as a capability today, and where it's headed.
Anatomie du prompt global — en français
Presentation in French · PDF · 7 slides. Created as part of Fred’s work to optimize and improve prompt management, this presentation explores how prompts are structured and how platform instructions, agent roles, tools, and conversation context fit together. Prompt management involves subtle trade-offs: reducing token overhead while preserving clear instructions, consistent behavior, and maintainability. By sharing this work, Fred helps readers understand why these choices are difficult and why prompt design deserves careful engineering.
Key Topics
Focused reference pages, architecture first.
Platform Architecture →
Component topology, data flows, storage boundaries, network layout, and the security perimeter.
Feature Reference →
Concise, complete inventory of what Fred provides — ingestion formats, agent capabilities, execution contracts, chat UI, and the governance model.
Erasure you can prove →
How Fred deletes a conversation — completely, provably, and on the team's own terms.
Deploying on GCP / GKE →
A typical GKE deployment: workloads, managed services, Workload Identity, TLS, secrets, GitOps.
Bootstrap, capabilities & teams →
From zero admins to a running platform: one-shot bootstrap, deliberate capability enablement.
Add your agent →
Build a custom agent with fred-sdk and fred-runtime, run it locally, enroll it with the control plane.
Access model →
Keycloak identity, ReBAC (OpenFGA) enforcement, the display-only RBAC catalog, per-endpoint checks.
Policy-based LLM routing →
How Fred resolves the effective model for an agent — capability, context, profiles, rules.
Durable execution →
Why ingestion, evaluation, and agent execution pair a fast API with a Temporal-backed workflow.
Observability & KPIs →
What Fred exports to Prometheus, where, and where it deliberately doesn't.