Applies the four-reviewer pass over the forge/dashboard-actions work. The through-line is that several capabilities were wired but inert, and one of them could cost real work. DUPLICATE WORK. "Closes #N" was dropped from PR bodies in favour of LinkBranch, which is sidebar linkage on Gitea and a no-op stub on GitHub -- neither closes anything, and nothing else did either. So a merged task left its issue open, labelled and assigned; the next poll re-enqueued it, and once the dashboard's Clear removed the task row that became a second implementation and a second PR for work already merged. reconcilePRs now closes the issue on merge, and a rejected task closes its issue too. A PR closed without merging deliberately leaves the issue open: that usually means "try again". MAX RETRIES. Configured, defaulted and reported by /api/config, and enforced nowhere -- markDead was unreachable and EffectiveMaxRetries had no callers, so the dashboard's retry incremented retry_count forever. The retry action now retires a task to dead at the cap. DEAD ON ARRIVAL. Tasks() never selected plan or source, so the "Decision required" panel could not render and chat tasks -- whose issue number is synthetic -- linked to forge issues that do not exist. COSMETIC FAILURE, FATAL RESULT. OpenPR failed the whole stage when LinkBranch errored, so the pull request was never opened and the task parked; a retry re-links the same branch, which Gitea answers with a conflict, so every retry parked again. It is best-effort now. INVISIBLE ACTIONS. Approve/retry/reject wrote store rows and nothing else, so the timeline and activity stream never showed that a human had intervened. They emit events now. Also: store write failures answer 500 rather than 409; the dashboard surfaces the server's actual reason instead of "409 Conflict"; action buttons stop propagation (acting on a task no longer also expands it), disable in flight, keep per-task errors instead of one shared string, and confirm before rejecting. Removed: the maybeRetryParked shim and markDead with their now-vacuous tests, WaitingTasks (no consumers; its tests were rewritten to assert the same routing through TaskByID), write-only Daemon.Runtime, and dir/dirs response bloat from /api/skills. skillSource checked workDir before sharedDir, mislabelling every shared skill when the two are equal. DefaultRoots now dedupes roots differing only by a trailing separator. The air build rule claiming to keep ui/dist current never ran -- it required one file to satisfy include_dir and include_file together, and ui/ is excluded from the walk. Deleted with the comment that promised it. task test:ui joins task check and CI; it ran nowhere before, so the dashboard suite could rot silently. Docs corrected where they described behaviour that no longer exists: parks no longer comment on the issue, state labels no longer drive retries, and forgerpc's doc claimed four proxied methods while listing three of five.
152 lines
6.2 KiB
TOML
152 lines
6.2 KiB
TOML
# archied configuration. Copy to ~/.config/archie/config.toml.
|
|
# Secrets are NOT configured here directly: each one is a {engine, key}
|
|
# reference resolved through a secrets engine at startup. The built-in
|
|
# "env" engine reads key as an environment variable name. Custom engines
|
|
# (sops, bws, vault/openbao, ...) are Yaegi plugins loaded from a
|
|
# directory -- see internal/secret.
|
|
# secret_engine_dir = "~/.config/archie/secret-engines"
|
|
|
|
# work_dir = "~/.local/share/archie/work" # default
|
|
# db_path = "~/.local/share/archie/archie.db"
|
|
poll_interval = "60s"
|
|
label = "archie"
|
|
bot_user = "vectis-labs-archie"
|
|
diff_cap_lines = 400
|
|
# max_retries = 3 # how many operator-triggered retries a parked task gets
|
|
# before it moves to dead (global default; set per-repo to
|
|
# override). Enforced by the dashboard's retry action.
|
|
|
|
[forge]
|
|
type = "github"
|
|
host = "https://github.com"
|
|
token = { engine = "env", key = "ARCHIE_GITHUB_TOKEN" }
|
|
|
|
[dispatch]
|
|
trigger = "assignee" # "assignee", "label", or "either"
|
|
ack_reaction = "eyes" # "off" disables pickup reactions
|
|
|
|
[dispatch.labels]
|
|
queued = "archie:queued"
|
|
working = "archie:working"
|
|
waiting = "archie:waiting"
|
|
pr = "archie:pr"
|
|
parked = "archie:parked"
|
|
# dead = "archie:dead" # state label for a task that exhausted max_retries
|
|
|
|
[models] # role -> ai-sdk runtime ref (used by agent workflows, Phase 4b)
|
|
triage = "openai/gpt-5.4"
|
|
planner = "openai/gpt-5.4"
|
|
builder = "openai/gpt-5.4"
|
|
|
|
[providers.openai]
|
|
class = "openai"
|
|
api_key = { engine = "bws", key = "OPENAI_API_KEY" }
|
|
# api_key_env = "OPENAI_API_KEY" # backwards-compatible environment form
|
|
|
|
[providers.ollama]
|
|
class = "ollama"
|
|
|
|
# Optional interactive-chat catalog. /provider groups these references by
|
|
# their first path segment, and /model shows only the selected provider's
|
|
# entries. When omitted, Archie uses the distinct refs in [models].
|
|
#
|
|
# `operator` is the display name of the person this deployment assists. It is
|
|
# passed to the chat agent as runtime metadata, which means it is sent to the
|
|
# configured model provider on every turn -- leave it unset to tell the agent
|
|
# nothing about who is running it.
|
|
#
|
|
# `workspace` roots the chat agent's file and shell tools (read, write, edit,
|
|
# shell, grep, find). Unset, those tools are not registered at all. Set it and
|
|
# the agent can read, modify and execute inside that directory with the same
|
|
# reach the daemon process has -- choose it deliberately.
|
|
#
|
|
# `max_steps` caps the model/tool round-trips in a single chat turn
|
|
# (default 100). This is a task budget, not a safety limit: one question
|
|
# about a codebase easily costs a read, several greps and a few edits, and
|
|
# a turn that runs out stops mid-task looking like it gave up. To interrupt
|
|
# a turn, send /stop -- that cancels it immediately and clears anything
|
|
# queued behind it.
|
|
# [chat]
|
|
# operator = "Ada Lovelace"
|
|
# workspace = "~/.local/share/archie/workspace"
|
|
# max_steps = 100
|
|
# models = [
|
|
# "openai/gpt-5.6-sol",
|
|
# "openrouter/openai/gpt-5.6-sol",
|
|
# ]
|
|
|
|
# Optional self-hosted update adapter. Both values are argv arrays (never
|
|
# shell snippets). The check command must write a releaseupdate.Snapshot JSON
|
|
# document to stdout; configure the installer as well to show Approve/Defer.
|
|
# [chat.telegram]
|
|
# update_check_command = ["/usr/local/bin/archie-update-check"]
|
|
# update_install_command = ["/usr/local/bin/archie-update-install"]
|
|
|
|
# External MCP tools are launched and managed by Archie's MCP client.
|
|
# [[tools.mcp_servers]]
|
|
# name = "desktop-commander"
|
|
# transport = "stdio"
|
|
# command = "npx"
|
|
# args = ["-y", "@wonderwhy-er/desktop-commander@0.2.46", "--no-onboarding"]
|
|
# work_dir = "/path/to/workspace"
|
|
|
|
[agent]
|
|
mode = "inprocess" # "subprocess" runs each stage through archie-agent (stdin/stdout); "nats" sends stages over NATS
|
|
command = "archie-agent" # executable path for subprocess mode (not used in nats mode)
|
|
# Subprocess mode isolates the protocol and process lifecycle, not OS privileges.
|
|
# The worker still runs as the daemon user until the Docker sandbox phase lands.
|
|
# env = ["GOCACHE"] # extra environment variable names forwarded to the worker
|
|
|
|
[indexing] # workspace codesearch index; both paths default under work_dir
|
|
# index_dir = "/var/lib/archie/indexes" # codesearch sidecar files
|
|
# db_path = "/var/lib/archie/workspace-indexes.db" # index lifecycle metadata
|
|
|
|
[web] # observability dashboard (read-only, no auth -- keep it local/tailnet)
|
|
listen = "127.0.0.1:8484" # "off" disables
|
|
|
|
[notify] # optional webhook for human-in-the-loop moments (n8n -> email)
|
|
# webhook = "https://n8n.example.com/webhook/archie"
|
|
|
|
[budgets]
|
|
max_steps = 60
|
|
max_tokens = 400000
|
|
wall_clock = "30m"
|
|
gate_max_failures = 5
|
|
|
|
# NATS JetStream for distributed task queue. Remove or leave commented out
|
|
# to run in single-process SQLite mode.
|
|
# [nats]
|
|
# url = "nats://localhost:4222"
|
|
# token_env = "NATS_TOKEN" # optional; named variable must be set
|
|
|
|
# Docker sandbox execution of archie-agent. Requires [nats] and agent.mode = "nats".
|
|
# [containers]
|
|
# enabled = true
|
|
# image = "archie-agent:latest" # Docker image to run
|
|
# max_concurrency = 4 # max simultaneous tasks/containers
|
|
# max_uptime = "60m" # container lifetime cap
|
|
# volume_ttl = "72h" # max age for persistent repo volumes and Git clone caches
|
|
# pull_policy = "missing" # "missing" or "always"
|
|
|
|
[[repos]]
|
|
owner = "your-github-username"
|
|
name = "todo"
|
|
base = "main"
|
|
ecosystem = "go" # "go", "python", "node", "rust", or "custom"
|
|
# Override the ecosystem defaults when needed:
|
|
# preflight = [["go", "version"]]
|
|
# test_glob = "*_test.go"
|
|
protect = ["_templ.go"]
|
|
gate = [
|
|
["templ", "generate", "./components/"],
|
|
["go", "vet", "./..."],
|
|
["go", "build", "./..."],
|
|
["golangci-lint", "run"],
|
|
["go", "test", "./...", "-count=1"],
|
|
]
|
|
# allow_concurrent = true # let the daemon run multiple tasks for this repo at once instead of
|
|
# FIFO-serializing them; only safe when concurrent worktrees won't collide
|
|
# (e.g. different base branches or packages per task). Still bounded by
|
|
# [containers].max_concurrency globally.
|
|
# persistent_storage = true # reuse a per-repo Docker volume and Git object cache until volume_ttl
|