ThoughtFrame Core Platform Roadmap
The ThoughtFrame Core is the engine beneath every vertical — Frames, deterministic orchestration, and the emerging Deep Memory layer that gives long-running AI workflows structure, continuity, and context. Where the LMS sits at the surface, ThoughtFrame defines the internal machinery that makes adaptive, multi-step reasoning possible.
The platform has evolved through discrete phases: first the deterministic FSM-based Frame engine, then agentic crossover with LLM-driven branching, and later the invention of Frame Sessions as a durable execution container. The next milestone is the full integration of semantic, lineage-aware memory — transforming each Session into a long-horizon knowledge substrate that all Frames can leverage. This backbone will power every application built on ThoughtFrame, whether it’s learning, legal reasoning, compliance engines, or entirely new domains still to come.
ThoughtFrame Core Platform Roadmap
Frames, Deep Memory, Orchestration Engine
Phase 1: Core Engine (Completed)
Foundation
Goal: Establish a deterministic execution engine where Frames behave
predictably under all conditions, forming the mechanical foundation of the entire platform.
Outcome: A stable FSM-driven runtime capable of replaying state, handling
synchronous and asynchronous events, and maintaining strict control over how Frames evolve,
making higher-level orchestration possible.
- Deterministic FSM-driven Frame engine
- Sync + async event handling
- State progression with deterministic replay
- Stable property model and mutation rules
- Thread-safe orchestration primitives
Phase 2: Agentic Crossover (Completed)
Hybrid Intelligence
Goal: Allow Frames to incorporate LLM reasoning without losing
deterministic structure or predictable branching.
Outcome: Frames became hybrid agents—able to request model guidance, evaluate
responses, and translate them into controlled FSM transitions. ThoughtFrame gained the
ability to blend symbolic logic with probabilistic reasoning safely.
- LLM-driven decision branches within deterministic FSMs
- Structured handoff between Frames and models
- Controlled nondeterminism with guardrails
- Model outputs mapped into safe event transitions
Phase 3: Frame Sessions (Completed)
Deep Memory Container
Goal: Introduce a durable execution container capable of preserving state,
history, and long-horizon memory across ticks and external events, providing the backbone
that makes semantic reasoning possible.
Outcome: The creation of Frame Sessions, now integrated into the
platform as the structured Deep Memory layer. Sessions store lineage, replayable execution
traces, and all context required for long-running workflows. The underlying research and
implementation details are available on GitHub:
github.com/thoughtframe-ai
.
- Session as the long-horizon memory backbone
- Durable execution context across ticks and events
- Structured storage of lineage, state, and reasoning traces
- Deterministic rehydration of long-running workflows
Phase 4: Semantic Frame Sessions (In Progress)
Active Work
Goal: Transform Frame Sessions into a full semantic memory substrate capable
of summarizing, indexing, and understanding information before it is stored.
Outcome: A lineage-aware deep memory system: rolling summaries, embedded
metadata, and semantic ingestion pipelines that allow multi-frame reasoning and long-context
retrieval using structured, interpretable data.
- Semantic lineage: rolling summaries + metadata embeddings
- Structured deep memory for long-context reasoning
- Chunk → summary → supersummary pipelines
- Instant Indexer for semantic-first ingestion
- Unified substrate for multi-frame semantic coordination
Phase 5: LMS + UI + Ecosystem Reintegration
Active Work
Goal: Reconnect the evolving platform core to real user interfaces, learning
flows, and domain-specific applications in TestU, TRP, and ThoughtFrame Law.
Outcome: Frames drive personalization and adaptive logic within existing LMS
modules and UI components, enabling practical deployment of the Deep Memory engine across
training, tutoring, and legal workflows.
- Reconnecting Frames to TestU/TestReadyPro UI surfaces
- Frame-driven personalization inside existing LMS flows
- Shared Session Memory accessible from Pods + Modules
- Real-time adaptive tutoring interfaces
- Domain-specific Frame bundles (legal, training, compliance)
Phase 6: APIs, SDKs, and Python Bridge
Upcoming
Goal: Expose ThoughtFrame as a true platform—language-agnostic, API-driven,
and ready for external developers and enterprise integrations.
Outcome: A stable REST API, WebSocket transport, and unified Python SDK,
enabling orchestration, Frame control, and session manipulation from any modern stack. This
phase completes ThoughtFrame as an openly consumable engine.
- Structured REST API for full orchestration control
- Finalized WebSocket and RPC transport for Frames
- Stable Python SDK (
import thoughtframe as tf) - Language-agnostic integration patterns
- Plugin/extensibility model for external agents and tools