Structured observability for the iroh/SWIM layer: Aggregator, typed Event/Snapshot
types, Sink (NoopSink default), ProbeScheduler, process stats, and host/iroh/swim
introspection, plus the swactor-diag-collector, -postproc, and -iroh-relay binaries
that assemble and render per-run bundles. Generalizes the pipeline-parallel-inference
example to N stages and adds the topology-planner spec.
Signed-off-by: Zachery Aaron Shores-Chmielewski <zacheryasc@gmail.com>
Extend the single-GPU example into a two-node pipeline-parallel run that splits llama3.2:1b across two rented vast.ai GPUs and closes the autoregressive decode loop over iroh.
- topology: add linear-chain helpers where each stage derives its neighbours locally from `STAGE`/`NUM_STAGES`, registering `pp-entry`/`pp-exit`/`pp-stage-{i}` SWIM names
- messages: add `StageActivation` (bf16 hidden-state hand-off carrying position/seq_len/is_prefill) and `NextToken` (sampled-token feedback with a `done` flag) that close the autoregressive loop between stage 0 and stage 1
- stage_actor: add `Stage0Actor` (tokenize -> embed_and_forward -> prefill activation; decode_step on each NextToken) and `Stage1Actor` (forward_and_sample -> NextToken back; emit InferenceResponse on EOS/max_tokens)
- vastai: fork the client and add `create_pipeline_instances` (rents one instance per stage, threading `STAGE`/`NUM_STAGES`, best-effort destroys on partial failure) and `destroy_all_instances`
- pp_tinygrad_worker.py: per-stage worker slicing `model.blk[start:end]` in stub and real (GGUF) modes, plus new `pp_gpu_node`/`pp_smoke_run` binaries and ROADMAP/SPEC/TEST_SPEC docs
- reuse: build on the single-GPU example's iroh transport and process bridge unchanged; add actor/codec/topology/integration test suites
Signed-off-by: Zachery Aaron Shores-Chmielewski <zacheryasc@gmail.com>
Add a complete single-GPU distributed-inference example that rents a vast.ai GPU, boots a worker container, and runs a prompt end-to-end over iroh/SWIM.
- examples/single-gpu-inference: add the `single_gpu_inference` orchestrator binary that starts a local iroh node, waits for the remote gpu-node to register the `"inference"` SWIM name, then sends an `InferenceRequest` and prints the response
- examples/single-gpu-inference: add the `gpu_node` binary that joins the cluster via `SEED_ADDR`, spawns an `InferenceActor` over `tinygrad_worker.py`, and registers the `"inference"` bridge
- inference_actor: bridge swactor messaging to a Python child process via stdin/stdout JSON, with `ProcessBridge`/`RequestBridge` adapters that satisfy the single-`Incoming` actor constraint
- iroh_transport: add `IrohActorTransport` that sends `WireEnvelope`s over iroh QUIC uni-streams (connection-cached against early close), plus wire encode/decode and an inbound drain helper
- vastai: add a vast.ai REST client (`find_offer` with reliability/cuda/geo filters excluding CN, `create_instance`, `wait_for_running`, `destroy_instance`) parameterised by a mockable `base_url`
- worker/docs/tests: ship `tinygrad_worker.py`/`echo_worker.py` (newline-JSON, `--stub`/`--model` defaulting to llama3.2:1b), a Dockerfile, Makefile, SPEC, and actor/codec/cluster/integration/vastai test suites
Signed-off-by: Zachery Aaron Shores-Chmielewski <zacheryasc@gmail.com>
Make distribution and deployment more stable. Consolidate the logic for a generic swactor node.
Signed-off-by: Zachery Aaron Shores-Chmielewski <zacheryasc@gmail.com>
