smartcomputer-ai/lightspeed
57.0
Adequate · 9 November 2026
400.2k
lines of production code
Rust
with TypeScript
6
measurements over time
What this system is
Features
Add native Anthropic Messages API client
Introduces a new native client for the Anthropic Messages API within the llm-clients crate. This implementation supports standard API key authentication as well as OAuth bearer tokens, allows configuration of the base URL and API version via environment variables, and includes support for specific beta features such as MCP execution and thinking block binding controls.
crates/llm-clients/src/anthropic · high confidence
Agent profile registry and validation logic introduced
The \crates/profiles\ crate now provides the runtime registry and validation layer for agent profiles. It defines the \ProfileStore\ trait for creating, reading, updating (with optimistic concurrency via revision bumping), listing, and deleting profiles, alongside input extension traits that handle record creation and whole-document replacements while preserving identity and timestamps. Validation logic ensures required fields are present, timestamps are consistent, revisions are valid, and profile documents conform to metadata, retention, instructions, and environment attachment constraints.
crates/profiles · high confidence
Automated generation of MCP tool descriptors from API contracts
A new build script (generate-tools.mjs) has been added to automatically generate TypeScript tool descriptors for the Model Context Protocol (MCP) client. This script reads the platform's API contract (methods.json and api.schema.json), filters methods based on a tool-filter.json configuration, and outputs a validated tools.ts file. This ensures that the available MCP tools stay in sync with the backend API definitions and prevents name collisions or schema mismatches.
platform/configurator-mcp/scripts · high confidence
Company SSO integration with strict access controls and session management
The platform now supports Single Sign-On (SSO) for company users via OIDC, including a dedicated bootstrap mechanism for the initial platform administrator. Access is strictly governed by exact group membership claims, and sessions are bound to the specific OIDC issuer and an access version to ensure immediate invalidation upon role changes, suspensions, or issuer updates. The system also enforces a 'break-glass' emergency password login for designated admins when SSO is active, while explicitly blocking password-based authentication for company-identity users to prevent credential conflicts.
platform/web · high confidence
Embed build identity (version, git SHA, target) into binaries
The release-info crate now embeds build-time metadata directly into every released executable. A new build script captures the Git commit SHA, release version, Rust compiler version, and build target, exposing them as public constants (VERSION, GIT\_SHA, RUST\_VERSION, TARGET, LONG\_VERSION) and an envd\_targets helper. This allows users and tooling to inspect the exact build provenance of the binary at runtime, with robust handling for worktrees and packed refs.
crates/release-info · high confidence
Generated MCP tool definitions for workspace file reading and session management
The \platform/configurator-mcp/src/generated/tools.ts\ file now contains the generated TypeScript descriptors for the Lightspeed MCP tools. This includes the \vfs/workspaces/files/read\ tool for reading files from a workspace's current head, and the \session/start\ tool for creating or reopening sessions with configurable access, profiles, retention policies, and metadata. These definitions are automatically generated and provide the schema and type information required for interacting with the underlying platform services.
platform/configurator-mcp/src/generated · high confidence
Initial PostgreSQL schema for Lightspeed sessions, storage, and identity
This change introduces the foundational database schema for the Lightspeed platform, establishing the core tables required to manage sessions, content-addressed storage (CAS), virtual file systems (VFS), and identity. It defines the \universes\ table as the primary tenant boundary and \sessions\ to track lifecycle, visibility, and lineage (clones/forks). The schema also includes \vfs\_snapshots\ and \vfs\_workspaces\ for immutable file system state, \mcp\_servers\ for remote tool integration, and a comprehensive \auth\ substrate (\auth\_secrets\, \auth\_grants\, \auth\_clients\) to handle encrypted credentials and OAuth flows. Additional tables support environment compute (\environments\), agent profiles (\agent\_profiles\), API keys (\api\_keys\), and bot infrastructure (\bots\, \bot\_triggers\, \channel\_accounts\, \channel\_pairings\), providing the persistent state layer for the runtime.
crates/store-pg/migrations · high confidence
Initial release of the environment daemon with job, process, and identity management
The environment daemon is now available to manage execution environments. It introduces a JobManager for starting, listing, and reading jobs, and a ProcessManager for interactive process execution with PTY support and output retention. Daemon identity is handled via persistent Ed25519 keys stored in the state directory, enabling outbound registration with an environment gateway. Configuration is managed through CLI arguments and environment variables, supporting features like automatic upgrades, read-only filesystems, and secret environment variable redaction in process and job output.
crates/environment-daemon/src · high confidence
Introduce API projection layer for session and run views
A new \api-projection\ crate has been added to serve as the bridge between the internal reducer state and the stable client-facing API. This module translates internal event logs and blob storage into read-only \SessionView\ and \RunView\ structures, handling the projection of context entries, tool calls, and run summaries. It includes specific logic to decode native provider payloads (such as Anthropic text blocks and OpenAI responses) into displayable text while correctly handling binary media and preserving unicode integrity in previews.
crates/api-projection · high confidence
Introduce CAS-backed virtual filesystem models and snapshot helpers
The \crates/vfs\ crate now provides the core data structures and logic for a content-addressable storage (CAS) backed virtual filesystem. This includes deterministic path handling with strict validation (rejecting NUL bytes, empty components, and relative paths), immutable snapshot manifests that track file counts and byte totals, and helpers to create inline snapshots by writing file blobs and a manifest to a \BlobStore\. It also defines the catalog types for workspaces and snapshots, along with logic to resolve workspace attachments against available snapshots or live workspaces, ensuring that missing or unreadable targets are handled gracefully as unavailable projections rather than global failures.
crates/vfs · high confidence
Introduce Incus environment provider with public ingress support
This change adds the Incus provider implementation, enabling Lightspeed to manage development environments as Incus VMs. The provider supports both single-node and clustered Incus deployments, allowing administrators to define compute templates and configure network policies, including restricted egress CIDRs. A new stateless edge proxy is introduced to handle public ingress, enabling on-demand TLS authorization and proxying of HTTP and WebSocket traffic to running environments via stable, tenant-scoped hostnames. The provider also includes a relay service for on-demand data connections to the environment daemon and ships with a systemd service and build script for deploying the environment image.
crates/environment-provider-incus/src · high confidence
Introduce Lightspeed CLI with chat, authentication, and administration commands
The CLI now includes a new \lightspeed\ command set for interacting with the Lightspeed agent runtime. Users can start, resume, and manage interactive chat sessions via a terminal UI or JSON output, including model selection, workspace attachment, and tool call visibility. Authentication is managed through \lightspeed auth\, supporting OAuth client configuration, GitHub App provider setup, and credential binding. Administration commands (\lightspeed admin\) allow listing and switching between universes, managing API keys, and configuring runtime environments.
crates/cli · high confidence
Introduce Lightspeed Configurator MCP server
This change adds a new Model Context Protocol (MCP) server that exposes Lightspeed API methods as tools. It provides a configurable HTTP endpoint (defaulting to port 18081) that accepts MCP requests, authenticates callers via Bearer API keys (with optional universe and actor context), and dynamically exposes tools based on the caller's access groups. The server validates tool arguments using JSON Schema, enforces request size limits, and safely redacts sensitive credentials from error responses before returning results to clients.
platform/configurator-mcp/src · high confidence
Introduce Lightspeed JSON-RPC API contract and Rust access model
The \crates/api\ crate now defines the Lightspeed JSON-RPC API contract, including the \initialize\ handshake, session lifecycle methods (start, read, list, config, rename, metadata, retention, close, delete, share), event streaming, and workspace file access. It introduces a structured access model where API keys are scoped to either a specific universe or the entire deployment, and methods are grouped (e.g., \session\, \vfs\, \bots\, \channels\) to control key permissions. The crate also exposes the wire schemas for authentication grants, bot identities, and chat channel accounts, and provides a build tool to export the API reference, JSON schemas, and method definitions.
crates/api · high confidence
Introduce Lightspeed TypeScript SDK
The TypeScript client directory now contains the initial release of the Lightspeed TypeScript SDK (\@lightspeed-ai/sdk\). This new package provides a typed JSON-RPC client for interacting with the Lightspeed API, featuring generated types and method maps derived from the Rust backend contract. It includes workflow helpers for Temporal integration, allowing developers to build automated sessions and handle events like run completion and tool invocations. The release also includes a comprehensive test suite to validate the client's behavior and the generated workflow contract vectors.
clients · high confidence
Introduce Lightspeed Workflow Contract and Temporal orchestration infrastructure
This change establishes the foundational Temporal workflow runtime for the Lightspeed platform. It introduces a formal workflow contract (exported via \export-workflow-contract\ to \workflow-contract.md\, \workflow.json\, and \workflow.schema.json\) that defines the \deliver\_emission\ signal transport, emission ID framing, and producer authorization rules for cross-workflow communication. The crate implements the core \AgentSessionWorkflow\ and \EnvironmentJobWorkflow\ along with a comprehensive set of activities (e.g., \create\_or\_load\_session\, \llm\_generate\, \tool\_invoke\_batch\) and deterministic session log reduction logic (\rehydrate.rs\) to ensure reliable state recovery. Configuration constants define bounded timeouts, retry policies, and concurrency limits for tool and LLM activities, while helper utilities manage Temporal client connections and session preparation intents.
crates/temporal-workflow · high confidence
Introduce Lightspeed prompt instruction assembly and VFS root resolution
The system now assembles prompt instructions using a new modular tooling layer in \crates/tools/src/prompts\. This includes an assembler that builds prompt instruction sets from VFS roots with size limits and reporting, an environment prompt module that scans and publishes filesystem observations, and a VFS resolver that discovers prompt roots from workspace attachments (specifically \.lightspeed/prompts\ and \.agents/prompts\ directories). The implementation uses content-addressable storage (CAS) for blob management and defines specific schema versions for prompt instructions and source fingerprints under the \lightspeed\ namespace.
crates/tools/src/prompts · high confidence
Introduce TypeScript Lightspeed client and workflow runtime libraries
This change adds the initial TypeScript implementation of the Lightspeed client and workflow runtime. The \LightspeedClient\ class provides a JSON-RPC interface for interacting with the Lightspeed API, including methods for starting and managing sessions, reading events, and handling RPC errors. Additionally, generated types and method definitions are included to ensure type safety, and a workflow module is introduced to handle emission envelopes, workflow IDs, and channel connector activities, enabling integration with the underlying workflow system.
clients/typescript · high confidence
Introduce core authentication substrate with API keys, OAuth, and provider integrations
The \crates/auth\ module now provides the foundational identity and access layer, introducing structured API key management (minting, hashing, and scoping), a generic OAuth authorization-code flow service with PKCE, and a token broker that handles audience enforcement and single-flight token renewal. It also adds specific drivers for GitHub App installation tokens and MCP OAuth, along with the underlying grant, secret, and in-memory store abstractions that concrete persistence adapters will implement.
crates/auth/src · high confidence
Introduce environment registry with provider bindings and idempotent provisioning
The \crates/environments\ library now provides the core domain contracts and an in-memory implementation for managing environment providers, universe-scoped provider bindings, and environment incarnations. Users can register providers and bind them to universes with optimistic concurrency control (revision-based updates), and provision environments using deterministic request IDs that ensure idempotency on retries. The system also supports adopting existing environments from legacy sources and enforces constraints such as requiring all bindings to be removed before a provider can be deleted.
crates/environments · high confidence
Introduce environment selection and durable job execution tools
Users can now manage multiple attached environments within a session using new control tools: \environment\_list\ to view attachments, \environment\_activate\ and \environment\_deactivate\ to switch the active environment, and \environment\_read\ to inspect live status and access. The system also introduces durable job capabilities via \job\_submit\ and \job\_run\, allowing users to start, read, and manage background tasks with dependency support and timeouts. Additionally, process execution is exposed through \run\_process\ and \continue\_process\, enabling command execution with input/output handling and signal management. These tools are backed by a new environment catalog and VFS projection system that publishes stable access grants and file route details to the model.
crates/tools/src/environment · high confidence
Introduce inline tool runtime and tool catalog infrastructure
Added a new \crates/tools/src/runtime\ module that provides the \InlineToolRuntime\ for executing tool calls, a \ToolCatalog\ for managing tool bindings, and a \ToolTarget\ for resolving tools based on the specific model and provider. This infrastructure enables the system to resolve and invoke built-in tools (such as web fetch and file system operations) with support for environment contexts, VFS attachments, and media attachments in tool outputs.
crates/tools/src/runtime · high confidence
Introduce provider-native LLM client primitives for Lightspeed
This change introduces the \llm-clients\ crate, a lower-level library that manages provider HTTP transport, native API request/response handling, stream parsing, and error classification for Lightspeed. It adds specific content projection logic to extract visible text from Anthropic and OpenAI responses (including reasoning and refusal states) while preserving encrypted or signature-bound data for replay. The crate also defines a shared error taxonomy (e.g., \ProviderHttpError\, \TransportError\) with automatic retryability classification based on provider failure kinds, and supports per-request authentication overrides via API keys or OAuth bearer tokens.
crates/llm-clients/src · high confidence
Introduce structured filesystem abstraction with scoped access policies
The \crates/tools/src/fs\ module now provides a unified filesystem interface (\FileSystem\) that supports multiple backends, including a local host implementation, a deterministic in-memory version for testing, and VFS adapters for snapshots and workspaces. Access is governed by a new \FileAccessPolicy\ enum that distinguishes between full and scoped read/write or read-only modes, allowing tools to safely restrict file operations to specific root directories. This change introduces a \ScopedLocalFileSystem\ and \ScopedFileSystem\ wrappers to enforce these boundaries, ensuring that filesystem tools operate within defined limits rather than having unrestricted host access.
crates/tools/src/fs · high confidence
Introduce structured release packaging and Dockerized service images
This change establishes the release infrastructure for Lightspeed v0.4.0 by adding Dockerfiles for the runtime, platform, platform workers, and configurator MCP, alongside a formal release manifest schema. The runtime is now distributed as a static musl binary to ensure compatibility across Linux environments, while the platform services are containerized with pinned Node.js and Rust toolchain versions. The release manifest schema enforces strict versioning, protocol alignment, and artifact integrity for all components.
release · high confidence
Introduces MCP server registry and deterministic tool search
The \crates/mcp\ crate now provides the core contracts for managing remote MCP server catalogs and discovering tools. It defines the \McpServerId\ type and registry error handling, along with models for discovered tools and their annotations. A new deterministic search module allows users to find tools by name, description, and input schema parameters using a scoring algorithm that handles camel-case normalization, typos, and partial matches.
crates/mcp · high confidence
Introduces environment-protocol v2 with controller and data-plane definitions
The \environment-protocol\ crate now defines protocol version 2, introducing a structured controller plane for target lifecycle (listing, creating, adopting, and managing ingress) and a comprehensive data plane for file operations, job execution, and process management. Users can now interact with environments via explicit JSON-RPC methods for filesystem reads/writes, bounded text searches, durable job scheduling with dependencies, and process I/O with cursor-based output tracking. The protocol also supports large file transfers through bounded inline copies and multi-step transfer sessions, along with idle power reporting and portable filesystem inventories for cataloging. Also in this area: Introduce ephemeral JavaScript execution engine for composing session tools.
crates/environment-protocol/tests · high confidence
Introduces structured built-in tool surfaces and attachment handling
The \crates/tools/src\ module now defines a structured surface layer for built-in tools, presenting the same underlying filesystem and environment operations through provider-specific interfaces (Canonical, Codex-like, and Claude Code-like). This change introduces dedicated modules for these surfaces, allowing the tool descriptions, input schemas, and process output formatting to vary per provider while keeping the core logic shared. It also adds a new \attachments.rs\ module to manage file and media attachments, including logic to select linked attachments by handle and resolve immutable file references from the VFS or snapshots. Additionally, a \catalog.rs\ module is added to publish runtime-owned catalogs as immutable text with structured provenance, and a \concurrency.rs\ module defines the contract for promise-based concurrency tools (await, cancel, detach, sleep). Also in this area: Fix Anthropic tool schema compatibility by removing top-level combinators; Introduce ephemeral JavaScript execution engine for composing session tools.
crates/tools/src, crates/tools/src/builtin · high confidence
Native OpenAI API client library with audio, completions, and responses support
The \llm-clients\ crate now includes a native Rust implementation for the OpenAI API, providing dedicated clients for audio transcription (\audio.rs\), chat completions (\completions.rs\), and the Responses API (\responses.rs\). This library introduces a shared HTTP transport layer (\transport/\) that handles configuration via environment variables (e.g., \OPENAI\_API\_KEY\, \OPENAI\_BASE\_URL\), manages authentication headers, and includes a robust Server-Sent Events (SSE) parser for streaming responses. Users can now interact with OpenAI's endpoints directly through this typed client, supporting features like endpoint overrides, request timeouts, and sensitive header redaction.
crates/llm-clients/src/openai · high confidence
New LLM runtime adapters for Anthropic and OpenAI
The LLM runtime now includes dedicated adapters for the Anthropic Messages API and the OpenAI Chat Completions API, enabling the agent to interact with these providers using their native wire formats. These adapters handle the translation of provider-neutral requests into provider-specific messages, manage context compaction, and support native tool execution via the Model Context Protocol (MCP). The runtime also introduces robust media handling, normalizing images and PDFs to fit within provider limits, and implements a catalog system for managing context entries that can be updated without invalidating provider prompt caches.
crates/llm-runtime/src · high confidence
New Lightspeed agent evaluation harness for prompt-level testing
A new evaluation harness has been added to the \crates/eval\ crate, enabling users to run structured, prompt-level tests against the Lightspeed agent. The tool loads test cases from JSON files (located by default in \crates/eval/cases\), where each case defines setup instructions, expected outcomes (such as specific tool calls or file contents), and evaluation criteria. It supports running these cases against OpenAI and Anthropic providers, allowing for automated verification of agent behavior, tool usage, and response accuracy in a controlled environment.
crates/eval/src · high confidence
New automated release pipeline and verification scripts
The \scripts/release\ directory now contains a complete, automated release workflow. \build-dist.sh\ and \build-images.sh\ handle building the Rust binaries (including a static musl environment daemon), the TypeScript SDK, and Docker images for the runtime, configurator, and platform components. \prepare-version.mjs\ centralizes version management across the Rust workspace and metadata, while \create-manifest.mjs\ and \create-sbom.mjs\ generate a signed release manifest and a software bill of materials. The pipeline includes \publish-aliases.sh\ for tagging OCI artifacts, \smoke.sh\ and \smoke-images.sh\ for verifying binary integrity and container health, and \stage-package.mjs\ for preparing the npm client and configurator for distribution.
scripts/release · high confidence
New local development environment launcher and infrastructure management
Developers can now start the complete Lightspeed product with a single command (\./dev.sh\), which bootstraps dependencies, manages Docker Compose infrastructure (Postgres, pgAdmin, MinIO, Temporal), and runs editable Rust and TypeScript processes. The launcher supports six profiles (\full\, \platform\, \runtime\, \demo\, \docs\, \infra\) to tailor the local environment, handles CLI credential handoff and API key provisioning, and provides lifecycle commands (\status\, \stop\, \down\, \reset\) for managing the development stack.
scripts/dev · high confidence
New skill catalog and environment discovery system
The skills module now includes a comprehensive cataloging system that discovers and manages skills from both VFS workspace attachments and environment scans. Users benefit from automatic skill discovery in standard directories (\.agents/skills\, \.lightspeed/skills\), structured skill metadata parsing from \SKILL.md\ frontmatter, and a unified catalog publication mechanism that integrates skills into the runtime context. The system supports trust levels, scope definitions, and dependency tracking for discovered skills.
crates/tools/src/skills · high confidence
New test-support crate for in-process agent session harnesses
A new \crates/test-support\ crate has been added to provide fast, in-process harnesses for testing and evaluating the agent. It introduces a \SessionRunner\ that drives \CoreAgent\ sessions to quiescence, handling session storage, blob storage, and optional VFS workspace stores. The crate includes a \scripted\_tool\_id\ utility for resolving tool names in scripted tests, ensuring test models do not rely on provider-visible registry aliases. This support is explicitly not a production runtime and does not expose an \api::AgentApiService\ implementation.
crates/test-support · high confidence
New web\_fetch and web\_search tools with strict security and content handling
This change introduces two new AI model tools: \web\_fetch\ and \web\_search\. The \web\_fetch\ tool allows models to retrieve public HTTP/HTTPS URLs with strict SSRF protections (rejecting private IPs, localhost, and non-public schemes), configurable byte and character limits, and automatic text extraction from HTML, JSON, and plain text. The \web\_search\ tool provides a provider-neutral interface for web search capabilities, specifically implementing configuration for OpenAI's hosted search tool (supporting cached/live modes, domain filters, and user location). These tools are designed to be safe for untrusted web content retrieval and search operations.
crates/tools/src/web · high confidence
Platform database schema initialization and evolution
The platform database schema is established with a baseline migration defining core tables for authentication and multi-tenancy, including \user\, \account\, \organization\, \member\, \session\, \invitation\, and \verification\. This baseline also introduces \universes\ and \universe\_setup\_installations\ to manage platform environments. Subsequent migrations extend this structure by adding a \features\ column to \universes\ for extensibility, and introducing \company\_identity\ support via new \user\ columns for OIDC integration (\oidc\_issuer\, \oidc\_subject\), access control (\company\_admitted\, \emergency\_admin\), and versioning (\access\_version\), alongside a new \identity\_audit\ table to track security events.
platform/db/migrations · high confidence
PostgreSQL persistence layer for core runtime entities
The \crates/store-pg\ crate introduces a comprehensive PostgreSQL-backed storage implementation for the application's core data models. This change adds persistent storage for access control and visibility rules, deployment-scoped API keys with rotation support, and encrypted secret management using AES-256-GCM. It also implements storage for bots, triggers, and their event logs, as well as chat channel accounts and pairings. Additionally, the crate provides a content-addressed storage (CAS) system with an in-memory blob cache, deduplication, and a background sweep mechanism for garbage collection, alongside deployment-level administration for universe statistics and creation.
crates/store-pg/src · high confidence
Removals
Removal of main application entry point
The primary executable source file (src/main.rs) has been deleted, removing the default "Hello, world!" program entry point from the project.
src · high confidence
Behavioural changes
1 commit (0 fixes) modifying crates/temporal-runtime, platform/backend
A change to existing behaviour in crates/temporal-runtime, platform/backend — 1 commit, 3 files: “Unify package names and upgrade Rust and JavaScript dependencies (\#111)” (refs \#111).
crates/temporal-runtime, platform/backend · low confidence · unverified
1 commit (0 fixes) modifying platform/web/e2e, platform/web/scripts
A change to existing behaviour in platform/web/e2e, platform/web/scripts — 1 commit, 4 files: “Add Playwright visual diffs of the demo build”.
platform/web/e2e, platform/web/scripts · medium confidence · unverified
11 commits (0 fixes) modifying crates/store-fs
A change to existing behaviour in crates/store-fs — 11 commits, 2 files: “blob refactor”; “p194 harness improvements”; “vfs store implementation”; and 8 more (refs \#124, \#89, \#111).
crates/store-fs · medium confidence · unverified
12 commits (0 fixes) modifying crates/tools/src/environment\_protocol
A change to existing behaviour in crates/tools/src/environment\_protocol — 12 commits, 4 files: “renamed host-protocol”; “p194 harness improvements”; “windows exec”; and 9 more (refs \#54, \#124, \#89, \#115, \#51, …).
_crates/tools/src/environment\protocol · medium confidence · unverified
14 commits (1 fix) modifying crates/temporal-runtime/src
A change to existing behaviour in crates/temporal-runtime/src — 14 commits (1 fix), 100 files: “p194 harness improvements”; “Improve session permissions, deletion, and workspace file management (\#114)”; “fixes for aks (\#128)”; and 11 more (refs \#114, \#128, \#124, \#115, \#111).
crates/temporal-runtime/src · medium confidence · unverified
3 commits (0 fixes) modifying crates/environment-daemon/src/exec
A change to existing behaviour in crates/environment-daemon/src/exec — 3 commits, 8 files: “Make spawn record lifecycle test reap its entire process group”; “windows exec”; “Environment protocol v4 and a simpler environment model (\#124)” (refs \#124).
crates/environment-daemon/src/exec · medium confidence · unverified
3 commits (0 fixes) modifying platform/db/src
A change to existing behaviour in platform/db/src — 3 commits, 4 files: “integrate first-party platform monorepo (\#45)”; “Add Entra workload identity login for PostgreSQL”; “Add company SSO and platform authorization (\#104)” (refs \#45, \#104).
platform/db/src · medium confidence · unverified
5 commits (0 fixes) modifying crates/tools/src/fs/apply\_patch
A change to existing behaviour in crates/tools/src/fs/apply\_patch — 5 commits, 5 files: “p194 harness improvements”; “windows exec and transfer”; “moved to 1.97.1”; and 2 more (refs \#124).
_crates/tools/src/fs/apply\patch · medium confidence · unverified
5 commits (0 fixes) modifying platform/backend/src
A change to existing behaviour in platform/backend/src — 5 commits, 28 files: “p194 harness improvements”; “Improve session permissions, deletion, and workspace file management (\#114)”; “Add Entra workload identity login for PostgreSQL”; and 2 more (refs \#114, \#111).
platform/backend/src · medium confidence · unverified
6 commits (1 fix) modifying crates/environment-protocol/src
A change to existing behaviour in crates/environment-protocol/src — 6 commits (1 fix), 18 files: “renamed host-protocol”; “session metadata”; “fix”; and 3 more (refs \#124).
crates/environment-protocol/src · medium confidence · unverified
7 commits (0 fixes) modifying crates/environment-client/src
A change to existing behaviour in crates/environment-client/src — 7 commits, 12 files: “renamed host-protocol”; “windows exec and transfer”; “Core loop: cancel, steer, queue (\#54)”; and 4 more (refs \#54, \#124, \#89, \#51).
crates/environment-client/src · medium confidence · unverified
7 commits (0 fixes) modifying crates/environment-daemon/src/filesystem
A change to existing behaviour in crates/environment-daemon/src/filesystem — 7 commits, 14 files: “windows exec”; “windows exec and transfer”; “Align filesystem capabilities and improve session transcripts and mobile UI (\#91)”; and 4 more (refs \#91, \#124, \#89, \#115, \#111).
crates/environment-daemon/src/filesystem · medium confidence · unverified
7 commits (0 fixes) modifying crates/environment-protocol/src/data
A change to existing behaviour in crates/environment-protocol/src/data — 7 commits, 15 files: “renamed host-protocol”; “windows exec”; “tool improvements”; and 4 more (refs \#124, \#89, \#51).
crates/environment-protocol/src/data · medium confidence · unverified
7 commits (0 fixes) modifying crates/harness
A change to existing behaviour in crates/harness — 7 commits, 53 files: “p194 harness improvements”; “windows exec and transfer”; “session api cleanup”; and 4 more (refs \#124, \#115, \#111).
crates/harness · medium confidence · unverified
7 commits (0 fixes) modifying platform/backend/src/routes
A change to existing behaviour in platform/backend/src/routes — 7 commits, 26 files: “p194 harness improvements”; “Improve session permissions, deletion, and workspace file management (\#114)”; “Add manual context compaction controls and prepare v0.4.0 (\#112)”; and 4 more (refs \#114, \#112, \#125, \#124, \#111).
platform/backend/src/routes · medium confidence · unverified
7 commits (0 fixes) modifying platform/db/src/schema
A change to existing behaviour in platform/db/src/schema — 7 commits, 3 files: “integrate first-party platform monorepo (\#45)”; “Improve session context reliability and workspace workflows (\#109)”; “Move bots and channels core into the Rust runtime (\#65)”; and 4 more (refs \#45, \#109, \#65, \#55, \#104).
platform/db/src/schema · medium confidence · unverified
Database migration validation and auth schema tooling
This change introduces new scripts in the platform database layer to ensure schema integrity and correct timezone handling. The \check-migrations.ts\ script validates that the database schema matches the expected structure (including specific tables like \universes\ and \user\, while ensuring core bot/channel tables are absent) and verifies that migration upgrades preserve local user data correctly. Additionally, \auth-codegen.ts\ provides a configuration for generating the Drizzle schema that mirrors the runtime authentication setup, and \patch-auth-schema.mjs\ automatically corrects timestamp columns to use timezone-aware types (\timestamptz\) after code generation.
platform/db/scripts · high confidence
Project renamed to Lightspeed with new developer tooling and documentation
The project has been renamed from 'Forge' to 'Lightspeed', reflected in the updated README, logo, and environment variable prefixes (e.g., LIGHTSPEED\_). A new \dev.sh\ launcher script and Makefile have been added to streamline local development, requiring Node.js 24+ and Docker, and supporting various startup profiles (full, platform, runtime, demo). The repository structure has been updated with a new \.dockerignore\, \tsconfig.json\, and \rust-toolchain.toml\ (Rust 1.99.0), while the \CLAUDE.md\ symlink was removed in favor of an expanded \AGENTS.md\ guide.
(repo-wide) · high confidence
Test coverage
2 commits adding/updating tests in crates/environment-provider-incus/tests; 5 commits adding/updating tests in crates/environment-daemon/tests; 6 commits adding/updating tests in crates/environment-client/tests; 8 commits adding/updating tests in crates/temporal-runtime/tests; Added live PostgreSQL integration tests for the store layer; Added live integration tests for LLM client providers; Added live integration tests for the Anthropic Messages adapter; Added test coverage for the MCP configurator; Added test support infrastructure for live provider and media tests.
Dependencies
Initial dependency manifests for Lightspeed monorepo
The project introduces a new monorepo structure with dependency manifests for the TypeScript SDK (\@lightspeed-ai/sdk\), the documentation site, and the platform backend, CLI, and web components. The TypeScript SDK and platform services now require Node.js 24 and utilize TypeScript 6, Vitest 5, and Zod 4. The Rust backend defines a new workspace with crates for the API, authentication, bots, channels, CLI, and environment management, relying on Rust edition 2024, Tokio 1, Axum 0.8, and Temporal SDK 1.
(dependencies) · high confidence
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
How this codebase got here
Score
- CAI 54 → 57 (+3.0)
- Rubric changed (rubric-2026.10.5 → rubric-2026.11.2) — scores are not directly comparable.
Lenses
- Code Health 66 → 67 (+1.2)
- Architecture 82 → 82 (+0.4)
- Maturity 51 → 80 (+29.6)
- Readiness 47 → 48 (+1.2)
- Security 72 → 74 (+1.7)
- Event Sourcing 100 → 100 (+0.0)
- Accessibility 61 → 62 (+1.5)
- Performance 100 → 60 (-40.0)
Resolved (232)
- AdminEnvironmentProvidersPage.BindUniverseDialog (cyclomatic 17) (platform/web/src/pages/AdminEnvironmentProvidersPage.tsx)
- Ambiguous parsing intent: parse suggests general deserialization (e.g. from wire format), while of suggests construction from a specific domain concept (a method name). In other types like Visibility, only parse is used. If of is just a convenience for parse, it creates redundancy. If they serve different purposes, the naming convention is inconsistent with the rest of the API (which uses parse for string-to-type conversion).
- ClassTooLong: Operation (crates/environment-daemon/src/filesystem/transfer/session.rs)
- DaemonArgs::into_config (cognitive 17) (crates/environment-daemon/src/config.rs)
- Documentation: contradicts the code (docs/site/README.md)
- Documentation: written for insiders (docs/roadmap/p169-immutable-artifacts.md)
- Documentation: written for insiders (docs/roadmap/p170-filesystem-working-directories-and-sources.md)
- Duplicated block (10 lines × 2) (crates/environment-daemon/src/jobs.rs)
- Duplicated block (10 lines × 2) (crates/harness/src/core/components/approval.rs)
- Duplicated block (10 lines × 2) (crates/harness/src/storage/blobs.rs)
- Duplicated block (10 lines × 2) (crates/temporal-runtime/src/gateway/http.rs)
- Duplicated block (10 lines × 2) (crates/temporal-runtime/src/gateway/service/session_jobs.rs)
- Duplicated block (10 lines × 2) (crates/temporal-runtime/src/universe.rs)
- Duplicated block (10–11 lines × 2) (crates/llm-clients/src/openai/responses.rs)
- Duplicated block (11 lines × 2) (crates/environment-daemon/src/filesystem.rs)
- Duplicated block (11 lines × 2) (crates/environment-daemon/src/jobs.rs)
- Duplicated block (11 lines × 2) (crates/environment-provider-incus/src/server.rs)
- Duplicated block (11 lines × 2) (crates/llm-clients/src/anthropic/messages.rs)
- Duplicated block (11 lines × 2) (crates/store-fs/src/session.rs)
- Duplicated block (11 lines × 2) (crates/store-fs/src/session.rs)
- …and 212 more
New (224)
- AdminEnvironmentProvidersPage.AdminEnvironmentProvidersPage (cyclomatic 17) (platform/web/src/pages/AdminEnvironmentProvidersPage.tsx)
- AdminEnvironmentProvidersPage.ProviderDialog (cognitive 21) (platform/web/src/pages/AdminEnvironmentProvidersPage.tsx)
- AdminEnvironmentProvidersPage.ProviderDialog (cyclomatic 23) (platform/web/src/pages/AdminEnvironmentProvidersPage.tsx)
- BrokerSecretResolver::resolve (cognitive 16) (crates/temporal-runtime/src/worker/secrets.rs)
- Capture::visit (cyclomatic 16) (crates/environment-daemon/src/filesystem/transfer/capture.rs)
- Change coupling: mod.rs ↔ session_tools.rs (crates/temporal-runtime/src/worker/activities/mod.rs)
- ChatSessionDriver::project_turns (cognitive 22) (crates/cli/src/chat/driver.rs)
- ClassTooLong: Context (crates/tools/src/environment_protocol/conformance.rs)
- ClassTooLong: Publish (crates/environment-daemon/src/filesystem/transfer/publish.rs)
- ClassTooLong: Reconciler (crates/temporal-runtime/src/environments/lifecycle.rs)
- ClassTooLong: SessionPreparationService (crates/temporal-runtime/src/gateway/service/session_preparation.rs)
- ClassTooLong: Shared (crates/environment-daemon/src/filesystem/transfer/mod.rs)
- ClassTooLong: SubagentService (crates/temporal-runtime/src/subagents.rs)
- Confinement::walk (cognitive 52) (crates/environment-daemon/src/filesystem/confinement.rs)
- Confinement::walk (cyclomatic 27) (crates/environment-daemon/src/filesystem/confinement.rs)
- ControlSession::run (cognitive 42) (crates/temporal-runtime/src/gateway/registration.rs)
- ControlSession::run (cyclomatic 25) (crates/temporal-runtime/src/gateway/registration.rs)
- CoreAgentProjector::project_tool_batches_for_run (cyclomatic 18) (crates/api-projection/src/lib.rs)
- Duplicated block (10 lines × 2) (crates/environment-daemon/src/filesystem/confinement.rs)
- Duplicated block (10 lines × 2) (crates/environment-provider-incus/src/server.rs)
- …and 204 more
Changes since last survey
- 33 commits — 30 feature/other, 3 fixes
By area
- (repo) — 6 commits
- platform/web — 6 commits
- (root) — 4 commits
- crates/store-pg — 3 commits
- crates/environment-daemon — 2 commits
- crates/environment-protocol — 2 commits
- crates/temporal-runtime — 2 commits
- crates/tools — 2 commits
- clients/typescript — 1 commit
- crates/api — 1 commit
- crates/test-support — 1 commit
- docs/roadmap — 1 commit
- release/notes — 1 commit
- scripts/release — 1 commit
Notable commits
- fix: Add tests for uncovered web UI pages; fix locale-dependent editor test
- fix: Fix Windows branch CI path conversion and roadmap wording
- fix: fixes for aks (#128)
- change: Add Entra workload identity login for PostgreSQL
- change: Add Playwright visual diffs of the demo build
- change: Add queued-message steering and freeform patch tools (#125)
- change: Add v0.6.0 release notes
- change: Align Windows tool fixtures and preserve bounded environment discovery
- change: Build and assemble Windows envd with coherent Linux releases
- change: Environment protocol v4 and a simpler environment model (#124)
- change: Make spawn record lifecycle test reap its entire process group
- change: Merge branch 'main' into uiux/testing
- change: Merge pull request #131 from smartcomputer-ai/harness-core-hardening-squashed
- change: Merge pull request #132 from smartcomputer-ai/release/0.6.0-notes
- change: Merge pull request #133 from smartcomputer-ai/uiux/testing
- change: Merge pull request #134 from smartcomputer-ai/azure-workload-id
- change: Merge pull request #135 from smartcomputer-ai/windows-native-envd
- change: Plan native Windows environment support
- change: Prepare Lightspeed v0.5.0 release (#127)
- change: Prepare Lightspeed v0.6.1 release
- …and 13 more
Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.
Survey your own repository
smartcomputer-ai/lightspeed was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.
About this page
- The score is its most recent published measurement, taken on 9 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
- Measured at commit 8e8545be4d833ede647cf0bd3aa555360426d38b — the exact code this score is about.
- Scored under rubric-2026.11.2 — the same rubric and the same method as every other entry in this index.
- Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.3+preprod-bb0cb800cd0f.