Skip to content

CHANGELOG.md

Latest commit

 

History

History
141 lines (103 loc) · 5.78 KB

CHANGELOG.md

File metadata and controls

141 lines (103 loc) · 5.78 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Entries for 0.1.x were reconstructed from git history when this file was added in 0.2.0.

[Unreleased]

Changed

  • Agent floor and pinned ref raised to 0.16.0. This is an adoption bump, not a capability requirement: nothing here depends on what 0.16.0 changed. That release renamed the engine's built-in bundle (amplifier-agent-behavioral-anchor → amplifier-agent-anchors) and dropped descriptive prose from the head of its system prompt. Neither reaches this application — the HTTP face, GET /v1/models, /v1/skills, /v1/modes, host-config handling passed through --host-config, PROTOCOL_VERSION, and the run/serve contract are all unchanged, and nothing here names that bundle. The floor moves so AGENT_PINNED_REF — the exact tag the launch-time self-heal installs — tracks the current engine instead of drifting a release behind.

    Existing installs below the floor are healed at next launch: the preflight forces a reinstall to v0.16.0 rather than trusting the agent's own update.

[0.4.0] — 2026-07-29

Changed

  • Agent floor and pinned ref raised to 0.12.0. --host-config has always been forwarded to amplifier-agent serve --config, but below 0.12.0 the agent's HTTP face dropped the file's provider.config block on every turn, so the flag silently did less under serve than it does under run. 0.12.0 is the first version where it actually applies. This also makes the agent's new debug.rawLlmPayloads key reachable from opencode: below 0.12.0 it is rejected as an unknown top-level key. No launcher change was needed — the passthrough was already correct; only the engine it points at had to catch up.

[0.3.0] — 2026-07-29

Added

  • E2E DTU profile forwards GITHUB_TOKEN, so the harness can exercise amplifier-agent's GitHub Copilot provider alongside anthropic. Optional — unset is fine and the stack runs anthropic-only. Requires amplifier-agent 0.11.0+, which namespaces Copilot's model ids (github-copilot/<model>) so they no longer collide with the native provider's. No launcher change was needed: Copilot models flow through the existing display_name → opencode name mapping and render as <Model> (GitHub) in the picker.
  • README documents GitHub Copilot as a provider option, including the export GITHUB_TOKEN=$(gh auth token) bridge.

Changed

  • Minimum required amplifier-agent version raised to 0.11.0, the first release that namespaces reseller model ids. Below it, Copilot's claude-sonnet-5 and claude-opus-5 are served under ids byte-identical to the native anthropic provider's and silently capture its traffic — so the Copilot setup this release documents is only safe at or above that floor. AGENT_PINNED_REF follows it to v0.11.0, meaning the launch-time auto-install/self-heal now targets that tag.

Fixed

  • __version__ was left at 0.1.3 when 0.2.0 shipped, so --version and doctor under-reported the installed adapter. Now tracks the version in pyproject.toml.

[0.2.0] — 2026-07-27

Added

  • Skills bridge. amplifier-agent skills become opencode slash commands (/<name>, running server-side). Written to ~/.config/opencode/command/ globally or <project>/.opencode/command/ with --project-dir.
  • Modes bridge. amplifier-agent modes become opencode primary agents, shown as <mode> (Amplifier) in the agent picker. Written to ~/.config/opencode/agent/ or <project>/.opencode/agent/.
  • Bridged files are ownership-tracked (.amplifier-generated-commands.json, .amplifier-generated-agents.json); a command or agent file you wrote yourself is never overwritten.
  • Name conflicts (a skill or mode name found in more than one discovery location) are reported at launch, showing which file runs and which are shadowed.
  • DTU-based end-to-end test framework driving the real opencode TUI (tests/e2e/).

Changed

  • Minimum required amplifier-agent version raised to 0.10.0. That is the first release shipping GET /v1/skills and GET /v1/modes, which both bridges read; against an older agent the bridges silently produce nothing.

Security

  • Skill/mode names are validated as safe bare filenames before they can become a command or agent file, closing a path-traversal vector.

[0.1.3] — 2026-07-21

Added

  • --version flag — reports amplifier-opencode's version plus the installed and required minimum amplifier-agent version.

[0.1.2] — 2026-07-21

Added

  • amplifier-opencode setup — install/heal the stack and connect a provider without launching.
  • Self-healing preflight on every launch (--yes, --no-bootstrap): installs/heals amplifier-agent and opencode, and walks through provider credential setup when none is configured. Auto-install targets a pinned amplifier-agent tag rather than a moving branch.
  • update --no-opencode — update amplifier-opencode and amplifier-agent while leaving opencode at its current version.

Changed

  • Minimum required amplifier-agent version raised to 0.9.3, the first release shipping auth set --stdin (used to hand the provider key to the agent off-argv).

[0.1.1] — 2026-07-14

Changed

  • Minimum required amplifier-agent version raised to 0.9.1, and amplifier-opencode update now cascades into updating the agent.

[0.1.0] — 2026-06-20

Added

  • Initial release. amplifier-opencode launches opencode against a local amplifier-agent HTTP server, materializing opencode config from GET /v1/models on every launch. Includes the doctor subcommand.