What's Changed

• feat(runtime): NUDGE_REPAIR_BUDGET — cumulative ceiling on repair-round spend by @NekomyaDev in #23
• feat: NTF v1.1 branch field + trace-view par lane coloring by @NekomyaDev in #24

Full Changelog: v1.2.0-alpha.2...v1.2.0-alpha.3

What's Changed

• feat: nudgec trace-view — local web UI for traces (v1.2) by @NekomyaDev in #21
• feat: nudgec trace-diff — compare two traces (v1.2) by @NekomyaDev in #22

Full Changelog: v1.2.0-alpha.1...v1.2.0-alpha.2

What's Changed

• feat: nudgec trace-view — local web UI for traces (v1.2) by @NekomyaDev in #21
• feat: nudgec trace-diff — compare two traces (v1.2) by @NekomyaDev in #22
• feat(runtime): NUDGE_REPAIR_BUDGET — cumulative ceiling on repair-round spend by @NekomyaDev in #23
• feat: NTF v1.1 branch field + trace-view par lane coloring by @NekomyaDev in #24
• feat: Anthropic + Mistral provider adapters by @NekomyaDev in #25
• feat: real SSE streaming for llm_stream by @NekomyaDev in #26
• fix: ```json fence tolerance in streamed prefix validation by @NekomyaDev in #27
• feat: TS runtime parity — par branch labels + Promise-leak fix by @NekomyaDev in #28
• feat: TS streaming — stream let emits rt.llmStream by @NekomyaDev in #29
• feat: DAP groundwork — nudgec debug steps through traces by @NekomyaDev in #30
• chore: P0 docs/version sync by @NekomyaDev in #31

Full Changelog: v1.0.1...v1.2.0

The compiler now lints your prompts. v1.2.0-alpha.1 ships Prompt Clippy (design §20): nudgec check warns on vague llm""" instructions, missing output contracts, schema fields never mentioned in the prompt, and prompts too long for the declared budget — surfaced in the editor via LSP (W0001–W0004).

Also included: the full v1.1 series — real providers (one stdlib-only adapter for OpenAI/Gemini/Groq/Ollama), prebuilt binaries, the VS Code extension on the Marketplace, LSP hover/definition/completion, and real MCP stdio transport.

Install: download the binary for your platform below, extract, add to PATH. Then:

nudgec check examples/research_agent.ndg   # types + effects + prompt lints

Full Changelog: v1.0.1...v1.2.0-alpha.1

Full Changelog: V1.0.0...v1.0.1

Nudge v1.0 — the typed, replayable, budget-aware language for LLM agents. Write the agent once, compile it to Python or TypeScript, replay its traces as tests at zero token cost.

Highlights

• Typed LLM calls — llm"""...""" with { schema: Report, budget: 0.04 USD, retry: 3 with repair }; schema validation with an automatic repair loop
• Trace-as-test — every run records a trace; test blocks replay it in CI at zero token cost (nudgec test)
• Budget walls — per-call and per-run USD budgets, enforced deterministically against the fake provider
• Hybrid replay — NUDGE_REPLAY_MODE=llm replays the LLM from a trace while tools run live
• Streaming — stream let with incremental schema validation and early abort
• Checkpoint/resume — agent/state blocks checkpoint every write; nudgec resume <run_id> continues crashed runs
• Reducer state — l | merge r CRDT-style joins (dict union, list append-dedup)
• Model routing — route{ cheap: "m1" when cond, strong: "m2" otherwise } with the chosen label in the trace
• Static cost report — nudgec cost counts llm call sites per fn (fake pricing, retry×repair worst case)
• Multi-server MCP routing — NUDGE_MCP_SERVERS registry, fail-fast on unknown servers
• Two backends — Python (full) and TypeScript (nudgec build-ts + shipped TS runtime)
• OTel export — NUDGE_OTEL=<path> writes every trace record as an OTel-shaped JSONL span
• Frozen trace format — v1 record schema frozen; nudgec trace-check validates conformance
• A2A export — nudgec a2a emits A2A agent cards from your agents/fns
• LSP — nudgec lsp serves editor diagnostics over stdio (dependency-free)

Numbers

• 97 tests green (CI on every push) · 29-line self-testing research agent example · zero compiler dependencies
• Design doc v1.13 frozen · full roadmap complete (v0.1 MVP → v0.2 → v0.3 → v0.4 → v1.0)

Quickstart

cargo build                              # the nudgec compiler
export PYTHONPATH=$PWD/runtime
nudgec check examples/research_agent.ndg
nudgec test  examples/research_agent.ndg # replay the committed trace — zero tokens