# Changelog All notable changes to this project will be documented in this file. The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html). --- ## [1.17.0] - 2026-06-03 ### Added - **Receipt Undo** - `mythos receipts undo ` replays a verified receipt in reverse. By default it previews the reversal; `--yes` applies it. Reversal actions run through the same security-policy review and SWD engine as any other write, and applying an undo produces its own verifiable receipt. Available in the SDK via `planUndo` / `executeUndo` / `undoReceipt`. - **Stats JSON Output** - `mythos stats --json` emits machine-readable budget analytics (sessions, tokens, cost, per-command and per-project breakdowns) for CI and automation. - **Coverage in CI** - Added a `test:coverage` script and a CI coverage report step (Node 22, Ubuntu leg). ### Security - **Drift-Gated Undo** - Undo only reverses a file when the working tree still matches what the receipt produced, so it never clobbers newer edits (override with `--force`). Tampered receipts (integrity-hash mismatch) are refused unless forced. - **Content-Free by Design** - Because receipts store hashes, not file content, undo fully reverses `CREATE`s (by deletion) and reports `MODIFY`/`DELETE` as not auto-reversible rather than guessing — surfacing a manual `git checkout` hint when the repo is under git. Undo can never write to a sensitive (blocked) path. ### Fixed - **Branding Consistency** - CLI `--help` description and the package description now say "Claude Opus 4.8", matching the rest of the project. - **Dream Duration Metric** - `dream` now records actual elapsed time instead of a hardcoded `0`. --- ## [1.16.0] - 2026-05-30 ### Added - **External Agent Action Schema** - Added `schemas/external-agent-actions.schema.json` and `mythos swd validate --file/--stdin --json` so outside agents can validate Mythos-compatible action envelopes before apply. - **Task Contracts** - JSON action envelopes can now include per-run `contract` boundaries with `allowedPaths`, `blockedPaths`, `requiredPaths`, and `expectedOutputs`. Contract failures stop before SWD writes anything. - **Run Outcomes Ledger** - Added local `.mythos/runs` records plus `mythos runs list` / `mythos runs show latest` for agent/model, receipt id, blocked paths, checks, rollback state, changed files, and task-contract status. - **Policy Suggestions** - Added `mythos policy suggest` to inspect repo structure and print suggested block/confirm guardrails without silently writing `.mythos/policy.json`. - **MCP Validation Tool** - Added read-only MCP `swd_validate`; MCP `swd_apply` now accepts task contracts and can opt out of run logging with `saveRun: false`. - **Compatibility Fixtures** - Added valid and invalid external-agent JSON examples for schema/contract validation. ### Security - **Fail-Closed Contracts** - Task contract checks run before security review, sandbox checks, receipts, run history, or filesystem mutation. - **No Raw Content in Run History** - Run records store paths, statuses, check summaries, receipt ids, and redacted errors, but not raw agent input or file contents. - **Read-Only Policy Inspection** - Policy suggestions are advisory only and never mutate project policy automatically. --- ## [1.15.0] - 2026-05-29 ### Added - **Isolated Runs** - `mythos swd apply --check ` and `--run-checks` test a batch in a throwaway copy of the project, run the checks there, and apply the same approved actions to the real working tree only if every check passes. Available over MCP `swd_apply` via `check[]` / `runChecks`. - **Policy-Declared Checks** - `.mythos/policy.json` now supports a validated, capped `checks` array of `{ name, command }` gates. Declaring checks never executes them; they run only with an explicit `--run-checks` opt-in. - **Sandbox Result in JSON** - `swd apply --json` output now includes a `sandbox` summary (per-check pass/fail and redacted output tail) for CI and automation. - **Receipt Markdown Format Alias** - `mythos receipts show --format markdown` now matches MCP `receipts_show` format naming, while preserving the existing `--markdown` and `--pr` flags. ### Security - **Fail-Closed Gating** - When isolated-run checks fail, the real working tree is never modified. The sandbox uses a 0700 temp dir, jails every write to the sandbox root (realpath + traversal rejection so a copied symlink cannot escape), excludes `.git`, caps mirrored file count, times out checks, and is always cleaned up. - **No Implicit Execution** - A cloned untrusted repository's `policy.json` cannot trigger command execution on its own; checks run only on explicit operator opt-in, and never during `--dry-run`. Check output is secret-redacted before display or receipts. ### Fixed - **Dry-Run Receipt Suppression** - CLI `mythos swd apply --dry-run` no longer writes SWD receipts by default, keeping preview mode side-effect-free. --- ## [1.14.0] - 2026-05-26 ### Added - **MCP Config Helper** - Added `mythos mcp config [generic|claude|cursor]` to print paste-ready stdio MCP client configuration. - **Project Policy File** - Added `.mythos/policy.json` for enforced repo-local SWD rules, including custom block/confirm globs, delete controls, operation allowlists, action batch limits, and write-size limits. - **PR-Ready Receipt Markdown** - Added `mythos receipts show latest --markdown`, `--pr`, and MCP `receipts_show` `format: "markdown"` for compact, paste-ready SWD receipt summaries in PR reviews. ### Security - **Repo-Local SWD Guardrails** - Project policy rules are enforced before filesystem mutation across `chat`, `run`, `swd apply`, and MCP `swd_apply`; malformed policy files fail closed. - **Nested Sensitive Path Guards** - Built-in SWD safety rules now block or require confirmation for sensitive and command-affecting files inside nested project directories, preserving monorepo protection for paths such as `apps/api/.env`, `services/web/Dockerfile`, and `packages/app/package.json`. ### Fixed - **Budget Limit Validation** - Invalid, non-finite, zero, or negative budget limits now fall back to safe defaults instead of producing meaningless percentage output. - **Provider Circuit Breaker Tuning** - Retryable provider failures now require consecutive exhausted requests before marking a provider degraded, avoiding a five-minute penalty from one transient 503-style event. - **Pricing Helper Documentation** - `getModelPricing` documentation now matches the implementation's conservative fallback behavior for unknown models. --- ## [1.13.0] - 2026-05-25 ### Added - **MCP Adapter for SWD** - Added `mythos mcp`, a local stdio Model Context Protocol server that exposes Mythos SWD, receipts, and skill inspection tools to MCP-compatible agent clients. - **MCP SWD Tools** - Added `swd_dry_run` and `swd_apply` tools that reuse the existing external-agent SWD boundary without calling a model provider, starting a daemon, or duplicating filesystem logic. - **MCP Inspection Tools** - Added read-only MCP tools for `receipts_list`, `receipts_show`, `receipts_verify`, `skills_list`, and `skills_check`. ### Security - **No HTTP Daemon** - The MCP adapter runs over stdio only, opens no local port, and keeps the same fail-closed sensitive path protection, rollback, and receipt behavior as `mythos swd apply`. --- ## [1.12.0] - 2026-05-24 ### Added - **External Agent SWD Interface** - Added `mythos swd apply` so external agents can submit structured file actions to Mythos without calling a model provider or requiring an Anthropic key. - **Model Free SWD Automation** - Added `--stdin`, `--file `, and `--json` support for machine readable external agent workflows. - **External Agent Receipt Metadata** - SWD receipts can now identify external agent/model sources for verified non-dry-run executions. ### Changed - **Provider Key Validation** - `mythos chat` and `mythos run` now require at least one configured provider key instead of requiring Anthropic specifically, preserving Anthropic as the recommended/default provider when present. - **README and SDK Documentation** - Documented the agent neutral SWD execution flow, security defaults, BYOK provider selection, and receipt behavior for external agent use. ### Security - **Fail Closed External Actions** - External SWD input is size limited, schema validated, constrained to safe project relative paths, and reviewed before filesystem mutation. - **Sensitive Path Protection** - External agent actions block `.env`, private keys, wallet files, `.git`, `.npmrc`, and secret-like paths by default. High impact command surface files and deletes require explicit `--allow-risky`. - **No Model Execution Boundary** - `mythos swd apply` does not call Anthropic, OpenAI, DeepSeek, provider fallback, memory compression, or test-healing; it only applies and verifies supplied file actions through SWD. --- ## [1.11.0] - 2026-05-22 ### Added - **`mythos learn` Command** - Added deterministic repo skill generation that creates `.mythos/skills/repo/SKILL.md` from local repo structure, docs, package scripts, CI workflows, public surfaces, and security-sensitive files without running project commands or calling a model. ### Changed - **Skill Onboarding** - Skills can now be bootstrapped from detected repository signals instead of requiring maintainers to write every rule pack from scratch. --- ## [1.10.0] - 2026-05-20 ### Added - **`mythos skills` Command** - Added first-class skill pack management through `mythos skills`, `mythos skills show `, `mythos skills new `, and `mythos skills check`. - **Project-Local Skill Packs** - Added `.mythos/skills//SKILL.md` support so repositories can ship their own Mythos operating rules without relying on a user's global setup. - **Global Skill Packs** - Preserved reusable user-global skills in `~/.mythos-router/skills//SKILL.md`, with project-local skills taking precedence when names overlap. - **Skill Receipt Metadata** - SWD receipts now record active skill ids, names, versions, and sources so verified edits can be reviewed with the rule packs that guided them. - **Skill Documentation and Examples** - Added a dedicated skills guide plus example `repo` and `security-review` skill packs. - **Skill SDK Helpers** - Exported skill loading, listing, validation, creation, and prompt-building helpers through the public SDK entry point. ### Changed - **Project Initialization** - `mythos init` now scaffolds and checks the project-local `.mythos/skills/` directory as part of repo onboarding. - **Skill Validation** - Skill checks now validate numeric limits, parse frontmatter arrays more consistently, and detect incompatibilities by either skill id or skill name. - **Receipt Privacy** - Receipt skill paths are stored only when they resolve inside the current project, avoiding accidental leakage of user-global or outside-project paths. --- ## [1.9.0] - 2026-05-19 ### Added - **`mythos run` Command** - Added one-shot prompt execution for tasks that do not need the interactive REPL. The command accepts any prompt, runs it through Mythos once, and exits. - **File and Stdin Prompt Sources** - `mythos run` can now read its prompt from a local file with `--file ` or from piped input with `--stdin`, making Mythos easier to use in scripts, task files, and editor workflows. - **Shared Chat/SWD Pipeline** - `run` reuses the existing chat session initialization, provider routing, SWD verification, receipts, memory logging, budget tracking, skills, and branch sandboxing instead of introducing a separate execution path. - **Bounded Run Defaults** - One-shot runs default to a smaller turn budget: one initial model turn, SWD correction turns, and optional test-healing turns only when `--test-cmd` is provided. - **Resume-Safe Execution** - `run` records metrics as its own command but does not overwrite the resumable session used by `mythos chat --resume`. - **`mythos init --check`** - Added a read-only setup check for environment, providers, `.mythosignore`, `MEMORY.md`, and the local skills directory without scaffolding or modifying files. ### Changed - **Command Help Coverage** - CLI smoke coverage now checks that built help output includes the `run` and `init` commands, verifies the `run --help` prompt-source options, and covers `init --check` as a no-write smoke path. --- ## [1.8.1] - 2026-05-17 ### Fixed - **SWD Rollback Drift Protection** — Rollback now uses the cached post-verification snapshot, preventing Mythos from overwriting external file changes made after verification. ### Changed - **CI Verification Gate** — Added a GitHub Actions step to run `node dist/cli.js verify --ci` against the locally built CLI. - **CI Hardening** — Tightened workflow permissions and install behavior. ### Security - **Local Data Disclosure** - Documented where Mythos stores local memory, receipts, resumable sessions, metrics, cache data, and skills so users can inspect or clear private project state. --- ## [1.8.0] — 2026-05-15 ### Added - **CI Verification Mode** — Added `mythos verify --ci`, a read-only GitHub CI mode for reviewing PR/diff changes before merge - **Generic PR Review** — `verify --ci` now works even when no Mythos receipts are present, checking high-impact repository changes in generic PR-review mode. - **Receipt-Aware CI Checks** — When Mythos receipts are changed under `.mythos/receipts/`, CI verifies receipt integrity and changed-file coverage. - **Execution-Surface Detection** — Added CI checks for `package.json` script changes, npm lifecycle hooks, GitHub Actions workflows, shell/deploy/Docker surfaces, `.env`/`.npmrc` paths, private-key-like files, and high-confidence secret patterns. - **CI Output Options** — Added `--strict`, `--json`, and `--base ` options for stricter CI policies, downstream tooling, and custom git base comparisons. - **CI Documentation** — Added `docs/CI.md` with GitHub Actions setup, exit behavior, examples, and maintainer notes. ### Changed - **Verify Command Extension** — Extended `mythos verify` with a dedicated CI path while keeping normal local verification behavior unchanged. - **Test-Healing Loop Refactor** — Refactored the test-healing loop in `src/commands/chat.ts` into smaller helper methods for maintainability, without changing existing chat/SWD behavior. ### Security - **No-AI CI Verification** — `verify --ci` does not call a model, use provider fallback, modify files, execute SWD actions, or write to `MEMORY.md`. - **Lifecycle Hook Review** — Newly added npm install lifecycle hooks such as `preinstall`, `install`, and `postinstall` are treated as high-severity CI findings. - **Execution-Surface Review** — Package scripts, workflows, shell/deploy files, and other high-impact repo surfaces are flagged for review before merge. - **Sensitive File Checks** — Added high-confidence checks for sensitive paths, private-key-like files, and secret-like material. --- ## [1.7.1] — 2026-05-13 ### Added - **Malformed Action Detection** — Mythos now warns when model output appears to include `[FILE_ACTION]` blocks but no valid actions can be parsed, making broken agent output easier to diagnose. - **Safety Regression Coverage** — Added tests covering receipt redaction, dry-run wording, and oversized write blocking. ### Changed - **Safer Receipt Output** — Receipt test-output tails are now limited to 500 characters and redact obvious API keys, tokens, and secrets before being stored locally. ### Fixed - **Large Write Protection** — Oversized `CREATE` and `MODIFY` actions are now blocked before touching disk, reducing the risk of unsafe full-file rewrites. --- ## [1.7.0] — 2026-05-11 ### Added - **SWD Trust Receipts** — Added persistent receipts for SWD runs, recording verified file outcomes, request summaries, provider/model metadata, token usage, git context, test status, and an integrity hash for later audit. - **`mythos receipts` Command** — Added receipt listing, inspection, and drift verification through `mythos receipts`, `mythos receipts show `, and `mythos receipts verify `, with `--json` output for automation. ### Fixed - **Machine-Readable JSON Output** — Terminal cursor restoration no longer contaminates redirected stdout, keeping `--json` output parseable in CI and shell pipelines. --- ## [1.6.1] — 2026-05-06 ### Fixed - **SWD Metadata Accuracy** — File hash metadata is now written only after successful, non-rolled-back SWD runs, preventing stale verification data after failed writes or rollbacks. - **Dream Metadata Preservation** — Memory compression now preserves `mythos:file` metadata, keeping SHA-256 drift detection intact after `mythos dream`. - **Test-Healing Drift Tracking** — Successful SWD fixes produced during test-healing now record file metadata for later verification. - **Anthropic Timeout Handling** — Anthropic requests now receive abort signals directly, improving watchdog timeout and fallback reliability during stalled streams. --- ## [1.6.0] — 2026-05-03 ### Added - **Cryptographic Drift Detection** — `mythos verify` now supports SHA-256 hash comparison for SWD-managed files. Hidden `mythos:file` metadata blocks in `MEMORY.md` store file state, allowing the verifier to detect manual edits, missing files, and unexpected content changes. - **CLI Smoke Tests** — Added `test/cli.test.ts` to validate the build lifecycle, built CLI execution, and `verify --dry-run` behavior in isolated temporary directories. - **Defensive Parser Limits** — Added a 250,000-character SWD block limit and stricter rejection of unsafe paths, including absolute paths, null bytes, traversal attempts, and oversized path values. ### Changed - **Terminal Visual System** — Added semantic `theme` and `icon` constants, improved mode badges, updated session/help/exit cards, and standardized success/warning/error/info colors across CLI output. - **Signal Handling** — Refined shutdown behavior so graceful exits return code `0`, while uncaught exceptions finalize safely and return code `1`. - **Mutation-Safe Cache Behavior** — Updated `ResponseCache` to bypass responses containing `[FILE_ACTION:]`, preventing file-mutating responses from being cached or replayed. - **Explicit `/clear` Confirmation** — Replaced interactive confirmation with `/clear confirm` to avoid nested readline conflicts in the chat REPL. ### Fixed - **SWD Parser Resilience** — Fixed handling of truncated or malformed `[FILE_ACTION:]` blocks to prevent parser stalls on incomplete model output. - **Verify Path Normalization** — Improved path normalization in `verify` to reduce false-positive drift results across relative path formats. --- ## [1.5.3] — 2026-05-01 ### Fixed - **Graceful Shutdown** — Removed competing `process.exit(0)` calls from `cli.ts` and `telemetry.ts` SIGINT handlers. Chat's `safeExit()` now fully owns the exit lifecycle, ensuring session save, metrics, and finalization always complete on Ctrl+C. - **SWD Content Preservation** — Replaced `.trim()` with precise boundary-newline removal in `parseActions()` so file content whitespace (leading spaces, trailing newlines, blank lines) is preserved exactly. - **Verify False Positives** — Fixed regex in `extractMentionedPaths()` that was including `chat:` entries and trailing semicolons as file paths, causing phantom "missing file" reports. - **Session Resume Turn Count** — Added `SessionBudget.restore()` so resumed sessions correctly hydrate token and turn counts instead of resetting to 1. ### Security - **Shell Safety** — `git.ts` uses `execFileSync` with argument arrays instead of string interpolation. --- ## [1.5.0] — 2026-04-30 ### Added - **`mythos init` Command** — Single-command project onboarding with environment validation (Node.js version, SQLite, Git), provider detection with actionable fix hints, and automatic scaffolding of `.mythosignore`, `MEMORY.md`, and skills directory. - **Project Scaffolding** — Automatically creates `.mythosignore` with sensible defaults to prevent accidental scanning of `node_modules` and other build artifacts. ### Changed - **Polished CLI Output** — Simplified internal jargon in user-facing logs (e.g., "derivative index" → "memory index"). - **Experimental Warning Suppression** — Automatically suppresses Node.js `ExperimentalWarning` for SQLite to maintain a premium, stable CLI experience. - **Provider Clarity** — Explicitly labels Anthropic as `required` and others as `optional` during initialization, providing clearer setup instructions. ### Fixed - **Dead Code Cleanup** — Removed orphan Next.js API routes that were accidentally included in the CLI repository. --- ## [1.4.0] — 2026-04-27 ### Added - **Session persistence & resume support** — Sessions are now atomically saved to `~/.mythos-router/sessions/latest.json` on exit. `mythos chat --resume` restores conversation history and budget state. - **Context window guard** — Automatically compresses the oldest portion of conversation history when approaching token limits using a low-effort summarization step. Prevents context overflow crashes during long sessions. ### Fixed - **CLI signal handling** — Improved handling of `SIGINT`, `SIGTERM`, and `uncaughtException`, ensuring terminal state is restored and sessions are safely persisted on exit or crash. - **Commander.js lifecycle** — Switched to `program.parseAsync()` to properly handle async command execution and prevent unhandled promise issues. - **Startup race condition** — Removed a dynamic import in the default help path that could cause banner inconsistencies in some environments. --- ## [1.3.1] — 2026-04-27 ### Added - **Developer Experience**: Added a `BACKLOG.md` to cleanly track and categorize deferred architectural decisions and future system robustness plans. ### Fixed - **CLI Robustness**: The terminal cursor is now safely restored if the CLI process crashes mid-animation, preventing the "invisible cursor" terminal bug. - **Provider Concurrency**: Fixed a decrement scoping bug in the `ProviderOrchestrator` ensuring `maxConcurrency` limits are strictly enforced even under parallel SDK usage. - **SQLite Compatibility**: Centralized `node:sqlite` loading into a single robust loader module, allowing the tool to fail gracefully instead of crashing on unsupported Node environments. ## [1.3.0] — 2026-04-26 ### Added - **Provider Observability Dashboard** — Added the `mythos providers` command. A real-time, terminal-based dashboard that surfaces EMA latency trends, success rates, circuit breaker recovery ETAs, and a live "Leader" score. Includes a zero-flicker `--watch` mode. - **SQLite Telemetry Backend** — Orchestration events are now persistently streamed to a dedicated `telemetry.db` using WAL pragmas and an asynchronous batching queue to guarantee zero hot-path blocking. - **Contextual Decision Tracing** — The engine no longer just routes; it explains *why*. Routing traces now explicitly capture task type (`chat`, `code`, `analysis`), input token buckets, and latency-vs-success-rate reasoning logs. - **Automated Retention Policies** — The telemetry engine aggressively self-prunes history, maintaining only the last 1,000 routing events and truncating error stack traces to prevent long-term database bloat. ## [1.2.1] ### Added - **Multi-Provider Orchestration Engine** — Decoupled the core application from the Anthropic SDK. The system now supports fallback routing, adaptive watchdogs, circuit breakers, and EMA-based performance scoring across multiple providers. - **OpenAI & DeepSeek Support** — Added a native, zero-dependency `fetch`-based provider (`OpenAIProvider`) to seamlessly support OpenAI and DeepSeek endpoints (including streaming reasoning content for `o1` and `DeepSeek-R1`). - **Skills Protocol** — Modular expert plugins via zero-dependency YAML frontmatter parsing. Skills (`-s `) can inject customized instructions, modify budget multipliers, and enforce deterministic provider selection. - **Deterministic Response Caching** — SDK utility for SQLite-backed response caching for pure-reasoning requests. Bypass rule strictly ensures file-mutating responses are never cached. - **Centralized Pricing Registry** — Unified token cost calculator across different providers, feeding exact financial data into the budget metrics. - **Auto-Healing TDD Loop** — Bounded, error-driven autonomy. Passing `--test-cmd` will automatically execute tests after a successful SWD mutation. If tests fail, the CLI intercepts `stderr`, truncates it, identifies TS/Runtime issues, and feeds it back to Claude for a self-healing iteration. - **TDD Anti-Thrashing Guards** — The orchestrator will automatically abort the healing loop if Claude attempts the exact same fix or if output remains identically broken, preventing runaway API costs. - **Contributor Covenant** — Added `CODE_OF_CONDUCT.md` to formally establish community standards. ### Fixed - **Stable Provider Fallback** — Fixed an edge case in the `ProviderOrchestrator` where identical EMA scores (e.g., at startup) could cause unpredictable provider routing due to JS `sort()` instability. The engine now explicitly respects user-configured priorities as a tie-breaker. ### Security - **CodeQL Integration** — Added GitHub CodeQL scanning badge to `README.md`. - **Dependency Audit** — Triaged and validated false-positive Socket.dev supply chain alerts for `@anthropic-ai/sdk`. --- ## [1.2.0] — 2026-04-23 ### Added - **SWDEngine v1 API** — Transactional filesystem execution kernel with `Plan → Snapshot → Execute → Verify → Rollback` lifecycle. Single entry point: `engine.run(actions)`. - **ChatUI Abstraction** — Decoupled chat session logic from the terminal via a `ChatUI` interface. `ChatSession` is now a pure orchestrator, fully testable and reusable outside the CLI. - **TerminalUI Implementation** — CLI-specific `ChatUI` adapter wrapping the Spinner and ANSI output. - **SWD Lifecycle Hooks** — Extensibility layer (`onAction`, `onVerify`, `onRollback`) allowing consumers to inject logging, telemetry, or custom UI into the engine. - **Rollback Auditability** — `SWDRunResult.rollbackErrors` field captures and reports rollback failures instead of silently swallowing them. - **`swd-cli.ts`** — Separated SWD terminal presentation (verification output, dry-run preview, verbose traces) from the pure execution kernel. - **Git Sandbox** — `ChatSession.setupSandbox()` for automated `mythos/` branch creation with nested-sandboxing protection. ### Changed - **SWD Kernel is now I/O-free** — `swd.ts` contains zero `console.log` calls. All presentation lives in `swd-cli.ts`. - **`validateApiKey()` throws instead of `process.exit(1)`** — library-safe error handling. - **SDK exports (`index.ts`) fully updated** — removed dead symbols (`runSWD`, `parseFileActions`, `snapshotFiles`), added `SWDEngine`, `parseActions`, `SessionBudget`, `ChatUI`, and all v1 types. ### Fixed - **🔴 Snapshot memoization bug** — `InternalSessionContext.getSnapshot('after')` was returning stale cached state on multi-action same-file scenarios. After snapshots now always re-read disk state. - **🔴 Broken `index.ts` exports** — SDK entry point was referencing pre-refactor symbols that no longer existed. --- ## [1.1.9] — 2026-04-22 ### Added - **Budget Analytics & Cost Profiling** — Persistent tracking of token usage and API costs across all sessions, projects, and commands. - **`mythos stats` Command** — New reporting engine for financial transparency. Aggregate costs by command, project, or time period (last N days). - **Global Metrics Store** — Local append-only JSON store in `~/.mythos-router/metrics.json` for cross-project financial auditing. - **Session Instrumentation** — Automated recording of chat sessions and memory compression (dream) events. --- ## [1.1.8] — 2026-04-20 ### Added - **Self-Healing Memory (V4)** — Re-architected memory system with a dual Authority/Derivative model. `MEMORY.md` remains the sole source of truth, backed by a rebuildable SQLite index. - **SQLite Derivative Index** — High-performance query acceleration layer using `node:sqlite`. - **FTS5 Smart Search** — Intelligent, ranked text retrieval via FTS5 virtual tables with `unicode61` tokenization. - **Integrity Signposting** — SHA-256 manifest hashing on startup ensuring zero drift between the authoritative log and the search index. - **Atomic Rebuilds** — Transactional reconstruction logic (`BEGIN/COMMIT`) to ensure index consistency even during hard crashes. ### Changed - **O(1) Append Protocol** — Optimized logging to use `appendFileSync` for better performance and durability under load. - **Hardened Test Suite** — Expanded testing to verify SQLite initialization, FTS5 search ranking, and recovery logic. --- ## [1.1.7] — 2026-04-19 ### Added - **Interactive Inline Diffs** — High-fidelity terminal previews for dry-run mode. Review exact line changes with ANSI coloring and line numbering before applying. - **Myers Diff Engine** — Implemented a zero-dependency, line-based shortest-edit-script algorithm in `src/diff.ts`. ### Changed - **SWD Protocol Upgrade** — Updated the "Capybara" system prompt to include the `CONTENT` field for 100% verifiability of file operations. - **Enhanced Regex Parsing** — Robust multi-line block extraction for complex code transfers. --- ## [1.1.6] — 2026-04-19 ### Added - **Atomic SWD Rollbacks** — Transactional filesystem safety. If any file action in a batch fails verification, the entire operation is reverted to its pristine state. - **Claude Opus 4.7 Support** — Official integration as the default `high` effort model. - **Adaptive Thinking Protocol** — Real-time streaming of model reasoning in the CLI REPL. - **Enhanced Dry-Run Previews** — Per-action confirmation prompts with detailed diff-style metadata. - **Adaptive Thinking Mode** — Full support in `client.ts` for thought-process streaming. ### Changed - Updated model identifiers for Claude Opus 4.7 compatibility. - Added SDK usage examples for programmatic integration. - Updated pricing constants and tokenization logic for Opus 4.7 compatibility. - Improved memory summarization "Dream" logic and session budget visualization. ### Fixed - Deduplicated internal `progressBar` utility. - `--effort` flag now validates input and warns on unrecognized values. - Improved path traversal detection in `resolveSafePath`. - Fixed memory leakage in long REPL sessions. --- ## [1.1.3] — 2026-04-17 ### Added - **Programmable SDK API** — Added the `src/index.ts` entry point and updated package module resolution. - **Exposed Modules** — Native export of `{ runSWD, streamMessage, snapshotFiles }` for external integration. - **SDK Documentation** — Integrated a new SDK Usage guide into the `README.md`. --- ## [1.1.2] — 2026-04-17 ### Added - **Multi-Model Orchestration** — Dynamic routing engine delegating tasks by effort (Opus 4.7 for Thinking, Sonnet 4.6 for Writing, Haiku 4.5 for Verifying). - **Dynamic CLI Badging** — Terminal now explicitly displays the exact model assigned to the current session. - **Protocol Tokenomics** — Added the official `TOKENOMICS.md`, formalizing the $MYTHOS Reasoning Tier Matrix. --- ## [1.1.1] — 2026-04-12 ### Fixed - **File Existence Accuracy** — Filesystem scanner recursively checks file existence for basic drift detection. - **True Dry-Run** — Fixed an issue where `MEMORY.md` was being created on disk even with the `--dry-run` flag. - **Memory Example** — Enriched the default `MEMORY.md` to reflect real sessions with file modifications. - **Codebase Polish** — Removed unused imports and obsolete Git-status checks. --- ## [1.1.0] — 2026-03-31 ### Added - **Financial Safety** — Budget limits and token tracker to help prevent bill-shock. - **Dry-Run Mode** — Preview all file operations with `[Y/n]` prompts before execution. - **Strict Write Discipline** — Enhanced verification logic for cleaner code. - **Codebase Verification** — Initial `verify` command implementation. --- ## [1.0.0] — 2026-03-29 ### Added - Initial release of mythos-router. - **Strict Write Discipline (SWD)** — pre/post filesystem snapshot verification. - **Adaptive Thinking** — Claude Opus with configurable effort levels. - **Self-Healing Memory** — `MEMORY.md` auto-logging with verification status. - **Correction Turns** — max 2 retries before yielding to human. - **Dream/Verify Commands** — memory compression and drift detection. [1.17.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.17.0 [1.16.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.16.0 [1.15.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.15.0 [1.14.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.14.0 [1.13.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.13.0 [1.12.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.12.0 [1.11.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.11.0 [1.10.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.10.0 [1.9.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.9.0 [1.8.1]: https://github.com/thewaltero/mythos-router/releases/tag/v1.8.1 [1.8.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.8.0 [1.7.1]: https://github.com/thewaltero/mythos-router/releases/tag/v1.7.1 [1.7.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.7.0 [1.6.1]: https://github.com/thewaltero/mythos-router/releases/tag/v1.6.1 [1.6.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.6.0 [1.5.3]: https://github.com/thewaltero/mythos-router/releases/tag/v1.5.3 [1.5.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.5.0 [1.4.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.4.0 [1.3.1]: https://github.com/thewaltero/mythos-router/releases/tag/v1.3.1 [1.3.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.3.0 [1.2.1]: https://github.com/thewaltero/mythos-router/releases/tag/v1.2.1 [1.2.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.2.0 [1.1.9]: https://github.com/thewaltero/mythos-router/releases/tag/v1.1.9 [1.1.8]: https://github.com/thewaltero/mythos-router/releases/tag/v1.1.8 [1.1.7]: https://github.com/thewaltero/mythos-router/releases/tag/v1.1.7 [1.1.6]: https://github.com/thewaltero/mythos-router/releases/tag/v1.1.6 [1.1.3]: https://github.com/thewaltero/mythos-router/releases/tag/v1.1.3 [1.1.2]: https://github.com/thewaltero/mythos-router/releases/tag/v1.1.2 [1.1.1]: https://github.com/thewaltero/mythos-router/releases/tag/v1.1.1 [1.1.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.1.0 [1.0.0]: https://github.com/thewaltero/mythos-router/releases/tag/v1.0.0