The AI runtime that never computes the same thing twice, and never loses its place.
Agent workflows burn money recomputing what they already know: the same system prompt tokenized ten thousand times, the same subtask answered again, an hour-long run lost to one crash at step 19. Continuum is a C++ execution engine that treats LLM calls and tensor ops as operators in one dataflow graph. Redundant work is cached at the runtime level, and a running workflow can be checkpointed to bytes, resumed in another process, or forked from any past step.
flowchart LR
P([prompt]) --> M{memo}
M -->|exact hit, 0 ms| R([result])
M --> S{semantic}
S -->|paraphrase hit, 0 ms| R
S --> T{trie prefix KV}
T -->|"shared prefix, ~99% fewer tokens sent"| B
T --> L{layer KV}
L -->|warm decode state| B[backend call]
B --> R
- 92.5% token reduction on a mixed 20-step agent workload against live Azure OpenAI.
- Zero-cost exact repeats. Memoized calls skip the backend entirely.
- Durable execution. Checkpoint, crash, resume, and time-travel fork, with deterministic replay.
- One graph for tokens and tensors. Azure, OpenAI, Anthropic, vLLM, libtorch, and MLX behind one IR.
python -m pip install continuum-aiKill an agent mid-run and finish it in a different process:
from continuum._native import DurableAgent
agent = DurableAgent()
agent.begin(["research the topic", "draft the report", "publish it"])
ckpt = agent.run_until_step(1) # bytes: graph + every value + KV cache state
# ... process dies here ...
revived = DurableAgent() # brand-new runtime
outputs = revived.resume_from(ckpt) # completes steps 3+ without redoing 1-2Rewind a finished run, edit one step, and replay the alternate timeline:
forked = DurableAgent.fork(ckpt, node_id, "write a haiku instead")
alternate = DurableAgent().resume_from(forked)See every reuse tier fire in one deterministic run:
PYTHONPATH=python python examples/01_reuse_stack.py # --trace for per-tier firing
PYTHONPATH=python python examples/02_durable_agent.py # checkpoint / crash / resume
PYTHONPATH=python python examples/03_time_travel_fork.py # rewind, edit, replayTen runnable scenarios (support-bot deflection, CI eval replay, agent-fleet
prefix sharing, spot-eviction resume, prompt A/B, hybrid tensor pipelines) are
indexed in examples/README.md.
Against a live Azure OpenAI backend, isolated per tier:
- Trie prefix KV cache: ~99% token reduction on a 3,000-char shared prefix.
- Memo table: 5/5 exact-repeat backend calls skipped.
- Mixed 20-step workflow: 92.5% token reduction, 4/20 backend calls eliminated.
Full tables, latency notes, and the scripts behind every number are in
docs/benchmarks.md, with raw data and reports under
benchmarks/.
- C++ execution engine with an IR interpreter and serializable checkpoints.
- Five-tier reuse stack: trie prefix KV cache, memo table, semantic cache, layer KV warm-start, memory graph recall.
- Durable execution: checkpoint a running workflow to bytes, resume in a fresh process with the KV cache included, or fork from a past step with an edited value.
- Session API with per-tier reuse policies and cross-session cache persistence.
- Backends: Azure OpenAI, OpenAI, Anthropic, vLLM shim, libtorch, MLX, and a deterministic FakeLLM for CI.
- v1 release hardening in progress.
- CIR schema lock with serialization conformance (
schema/cir.fbs). - Linux and macOS CI matrix with coverage gates and a fuzz workflow.
- PyPI packaging under
continuum-ai. Import path remainscontinuum.
- How Continuum fits with what you already use, plus what you can build and why it is a runtime, not a wrapper.
- Design docs: architecture, runtime model, cache semantics, IR.
- Benchmarks.
- Building the docs.
- Hosted API docs: Python, C++.
pip install pre-commit
pre-commit install
pre-commit run --all-files
pytest@software{continuum2026,
title = {Continuum: Unified Runtime for Token and Tensor Programs},
author = {Kamesh, Rithul and Contributors},
year = {2026},
url = {https://github.com/rithulkamesh/continuum},
version = {1.0.0}
}