Resolve the run workspace's environment with workenv (direnv, parent
.envrc, mise; same account rule as Claude chats) at run start and resume,
and apply it to Pi stage processes, a per-environment Pipelines-owned
OpenCode stage service (keyed by a hash of the change, reused across runs,
idle ones stopped after PIPELINES_OPENCODE_STAGE_IDLE_TIMEOUT, revived per
registration), delivery-gate checks and merges, plain-stage opencode CLI
calls and file-change capture git reads. Record the directory, .envrc
files, sources and variable names (never values) on the run detail; a
blocked .envrc becomes a run warning.
Each recorded artifact now carries a server group (deliverable or debug).
public/ files are deliverables; input/memory/history/output.md, session
exports and private/ scratch are debugging files. The app's Artifacts tab
lists deliverables and the Debug toggle reveals a separate Debugging files
group (older servers fall back to the filename rule); the TUI lists
deliverables first and marks debug rows; artifact search excludes debug
files unless include_debug / --debug-files is set.
Rewrite format.md around turn files, the output.md order (# outcome, edge word, # handoff, detail), checks, configs, human stages and the land stage delivery gate (sh -c in a fresh detached worktree, clean checkouts incl. untracked, 64 KB output cap, opencode only). Align review guidance across practices, verification, orchestration and the feedback-loop example. Fix loop-limit semantics in control.md against runner.go, budget_tokens sizing, read for large inputs, blocked-else-error and one stage naming set. Move dated verification logs to docs/verification-history.md and brokered chat, TUI keys and harness internals to docs/operator-reference.md; label operator-only references in SKILL.md.
The prompt now says output.md must start with # outcome and the outcome
word alone on the next line, then # handoff, then detail. Runs 241/243
followed the old "start with # handoff" wording and dropped the outcome.
When a completed turn lacks a valid outcome, the OpenCode/agent and Pi
clients send one short follow-up in the same session asking only for the
outcome line. An outcome-only reply goes at the top of the original report;
a still-invalid reply fails with the same validation error as before. The
follow-up's tokens accrue to the session usage. No outcome is inferred.
Missing tools/evidence now name blocked when declared, else error. The
parser tolerates a trailing period on the outcome word.
On a large Git workspace the 32 MiB retained-text budget was spent
alphabetically on files that never change, so small late edits were
recorded with no_patch. The stage-start snapshot now keeps text only for
files dirty or untracked at start; a clean tracked file keeps its hash and
stage-0 index blob id. When it changes or is deleted, the after capture
reads that blob in one `git cat-file --batch` and uses it only if its
SHA-256 matches the start read, so stage commits/resets do not move the
baseline and filters or assume-unchanged never yield a wrong patch. The
after capture shares unchanged files' representation instead of spending
budget on them again.
FileChange gains no_patch_reason (binary, file_size, budget, lines,
evidence_size, git_status, baseline). Non-Git workspaces keep their
behaviour.
OpenCode and Codex chats implement Detach (called by chatManager.Close)
instead of a parallel DetachSessions interface. Docs follow the Claude Code
host and the systemd KillMode=process unit. The opt-in live restart check
copies only the stage provider's login row into its scratch OpenCode
database, builds the CLI for skill install and removes its scratch tree.
A stopping server no longer interrupts harness sessions. Runs are canceled
with a shutdown cause, so stage clients leave admitted sessions running and
owned; start-up recovery re-adopts live OpenCode V2 stage sessions (wait,
read back the result, continue the graph), restarts unadmitted visits,
continues runs between stages, and falls back to interrupted with a reason
when the harness is gone. Chats with live work are recorded on shutdown and
reattached on start; turns that ended during the downtime are folded in
from history. The server-spawned V2 chat serve is now a detached owned
service. PIPELINES_SHUTDOWN_SESSIONS=stop restores the old behaviour.
A pipelines-claude-host process owns each claude -p CLI and its pipes,
spools its stream-json output with offsets and serves a private Unix
socket. The server detaches on shutdown and reattaches on start,
replaying the spool to rebuild turns, approvals, questions, background
tasks and the live model. Explicit close, idle release and CLI exit stop
the host; a crashed host falls back to --resume with a notice.
Stages now run on a detached Pipelines-owned opencode serve started with
OPENCODE_CONFIG_CONTENT carrying stage tool_output (16384 bytes / 400 lines)
and compaction.auto defaults, configurable through PIPELINES_OPENCODE_STAGE_*.
Chats keep their existing service and configuration; a server-spawned chat
serve can take PIPELINES_OPENCODE_CHAT_CONFIG. The service shares the user's
OpenCode config, credentials and session database, keeps a stable private
registration so retained references (recovery, interrupt, export) stay valid,
revives on the same port after a crash, replaces itself when idle after a
settings change, and is stopped on shutdown only when idle.
PIPELINES_OPENCODE_STAGE_SERVICE=shared restores the previous behaviour.
Every stage used to be told to read memory.md, the full output of every
earlier turn (19 KB average, 92 KB max), and paid for it on each model call.
memory.md is now essential memory, hard-capped at 4 KB and inlined in the
prompt: the task, the previous stage's # handoff section (or its truncated
output head), a one-line-per-stage outcome ledger and the server-captured
changed files. history.md keeps the old full memory on disk; read-capable
stages are pointed at it and input.md for on-demand detail, and it is never
inlined. The prompt asks each stage for a short # handoff section.
The server skip fallback for stages with a skipped edge now applies only
when every earlier visit recorded a complete file-change capture with no
changed files and no image artifact. Any change (no repo-specific UI path
or extension list), any partial/unavailable/missing capture, a human
decision or a checkpoint-retry prefix runs the model. auto_skip=false opts
a stage out. The fallback verdict names the server fallback.
Capture compares over-budget text by hash and, in Git, lists paths it
could not compare when their git status changes, marked no_patch, so a
large workspace no longer records files: [] for a real edit. Run detail
gains a warnings list for partial/unavailable captures, also printed by
inspect on stderr.
chat_sessions rows carry activity (turn_running, background_tasks, seq) from
the hub's own state, and chat_activity is pushed to every connection when a
session's turn starts or ends or its running background-task count changes.
A room whose last viewer left stays observed while background tasks run,
bounded by the unattended limits; a Claude Code process exit ends its tasks.
The template catalog is now Example 1-4: one stage, two stages (handoff),
feedback loop (runner-owned limits, retries, timeout) and human decision
(type=human approve/revise/abort with notify hooks). All read-only or leave
changes uncommitted; none commits, pushes or deploys.
A brand-new database (schema created by this Open) gets the examples once at
server start, bound to the configured default model and harness. A
server_settings marker records the decision, so removed examples never come
back and existing installs are marked without any pipeline being added or
rewritten.
Server: internal/discovery publishes a DNS-SD record (hashicorp/mdns) with
the port LAN clients dial (the web proxy via -mdns-port, which the
supervisor passes, or a non-loopback -addr) and TXT v, name, path, tls, auth.
On by default; -mdns=false or PIPELINES_MDNS=off disables it. Failures are
logged and never block startup; the advert is withdrawn on shutdown.
App: the Connect to Pipelines screen gets 'Find servers on this network'
behind a ServerDiscovery interface (Bonsoir: Android NSD with multicast lock,
iOS Bonjour with NSLocalNetworkUsageDescription and NSBonjourServices).
Tapping a server fills the address and connects; manual entry stays. Web has
no scan.
The composer's /new used to create a separate chat. It now sends
chat_new_session. The server starts a fresh native session on the chat's
own harness (OpenCode, Codex or Claude Code) and re-keys the
operator_sessions row in one transaction. The chat keeps its title,
directory, backend, colour, star, model and mode. The old row is kept
unlisted with replaced_by, and its native session stays resumable.
The reply and the chat_session_changed push carry previous_id, so every
client swaps the row in place and an open conversation follows it to the
fresh, empty transcript with a notice.
Chat hits sort by the chat's last use whatever field matched, and run,
step, field, artifact and pipeline hits sort by their own date, before
the result cap applies so the cap drops the oldest. Each kind's row scan
also walks newest first. Score still reports the match tier. The app
sorts both groups newest first as well.
A search with kind chat matches the operator-session registry's titles and
directories and the newest messages of recent transcripts through the chat
broker (bounded count, concurrency and one-second budget), answered off the
connection's read loop. The all-kind search is unchanged for older clients.
The app sends the chat search after its run search and shows a Chats group
with title, harness, directory and snippet; a chat hit opens that chat in the
Chat tab. The probe fixture answers chat searches over ?chats=N.
The chat information sheet gains a Session section: the harness's native
session id (copy), harness, directory, a verified resume command (copy)
and Export conversation (Markdown), Export JSON and, for Claude Code,
Export native session (the JSONL transcript).
Server: chat_sessions rows carry resume_command and native_export; new
chat_export command builds Markdown/JSON from the same paged history
chat_messages serves (bounded at 8 MiB), refuses non-operator sessions.
App: exports go through the artifact file bridge (web download, share
sheet on phones).
Usage frames carry optional usage.account {harness, source, plan, as_of,
windows[{id,label,used_percent,resets_at}], unsupported}. Claude Code quota
comes from its rate_limit_event stream lines and the initialize account
(plan), shared across chats on the same account. Codex reads
account/rateLimits/read once per connection (re-read when a chat opens after
5 min) and merges account/rateLimits/updated. OpenCode exposes no quota API
and says so. The app's chat info sheet shows an Account usage section with a
progress bar and reset countdown per window, Material and Cupertino.
Claude Code records subagent/peer reports, task notifications, reminders and
skill bodies as user-role entries. Classify them server-side (history and live
stream) by origin/isMeta/queued_command mode: reports become kind agent_report,
task notifications kind notice, harness meta is hidden, and the mid-turn
wrapper is unwrapped to the operator's text. The app draws a collapsed Agent
report card and a one-line system notice on the agent's side.
Claude Code chats now start with the chat directory's own environment:
direnv (allowed .envrc only, via direnv export json) and mise (mise env
--json), cached per directory and invalidated when config, watched files or
allow records change. A blocked .envrc is logged and shown in the chat as a
notice and never applied. When the directory chooses the Claude account
(CLAUDE_CONFIG_DIR, API key or token variables), the server's service token
and login-file fallbacks are skipped and inherited credentials dropped.
Transcripts in a directory's CLAUDE_CONFIG_DIR are found for history,
resume and after a restart. PWD is set to the chat directory.
Checks the CLI on the server's PATH (install command when missing), signs in
with the harness's own login only when needed (--login forces it), keeps the
OpenCode/Pi model pickers, and verifies with the server's check. Works on
native and Docker installs (Docker delegates to the container setup); --host
is a hidden compatibility alias. harness list shows all four harnesses'
installed/signed-in/configured/default status. ./install --opencode|--pi now
point to the command; ./install ends with the next-step hint.
Chat attention notifications (questions, approvals, finished replies) come
from a session's event stream, which the hub only read while a room existed;
detach destroyed the room when the last client left, so a turn that finished
or asked with the app closed notified nobody.
A room whose last client leaves now stays observing, unattended, while its
turn is in flight or it holds an unanswered question/approval, and is torn
down when the turn ends (or after 6h for a stuck turn); at most 16 such rooms.
At server start, recently used open sessions whose harness reports a running
turn (OpenCode /api/session/active, Codex thread/turns/list) are observed
again so their reply still notifies. Settings, dedupe and app_notifications
recording are unchanged.
Rewrite the README around the server, web/Android app and CLI, and
replace docs/install.md with the current native install (all-or-nothing
Docker web build, flags from setup-native.py, lifecycle commands, data
locations, password, upgrading). Add docs/harnesses.md and
docs/notifications.md; point ntfy.md at the new overview and fix its
topic name.
Server: list_directory returns a directory_listing (path, home, parent,
entries with is_dir/is_git_repo/hidden, truncated) confined to the server
user's home subtree after cleaning and resolving symlinks, refusing pseudo
filesystems and capping entries at 2000. Documented in docs/cli.md.
App: "Choose folder…" in the create dialog opens a picker (full-screen page
on phones, large dialog on wide windows, Material and Cupertino) with
breadcrumbs, filter, Show hidden, git badges, dimmed files, recent folders,
inline errors and "Start chat here". The probe app gains a folder fixture
for captures.
Claude Code (stream-json stdin) and Codex (turn/steer pinned to the active
turn) receive a prompt sent while a turn runs immediately. Other harnesses
(OpenCode) get a durable chat_prompt_queue row owned by the session: it
survives the last client detaching and a server restart, is delivered in
order when the turn ends with or without a client, and is deduplicated by
operation id. chat_subscribe carries the queue, chat_queue frames report
delivered/removed/failed changes, and chat_unqueue removes a waiting prompt.
An exhausted export recovery job left its run parked forever in the
export wait, reserving the workspace and refusing later scheduled runs
(runs 221/226: OpenCode sessions above the 8 MiB export bound are
always truncated, so every retry stays partial).
A gate is now releasable when its job is exhausted: the visit keeps its
partial/unavailable evidence and diagnostic (still repairable), the run
continues to its next stage or terminal, and the workspace is freed.
Gates older than 15 minutes with a pending, lease-expired or missing job
are exhausted by a bounded wait. The worker reconciles on its first
tick after startup and every 30 seconds, so existing wedged runs recover
after a restart. Truncation diagnostics now name the size bound.
Record every notification published to ntfy (run outbox, stage notify
hooks, chat question/approval/reply) in a persisted app_notifications
outbox and stream it to subscribe_notifications clients with an after
cursor; clear_notifications and ack_attention clear across devices, as do
runs leaving waiting, deleted runs and answered chat questions. Add the
notify.ntfy_enabled setting (default on) to stop only the ntfy publish.
Every run summary carries size_bytes (artifact tree, server-created Pi
session directories and an estimate of the run's database rows), cached in
pipeline_runs.size_bytes, invalidated by triggers, measured when a run
finishes and backfilled in the background. delete_runs removes terminal
runs with every scoped row and server-owned file, refuses active ones with
a reason, and pushes runs_removed to subscribers. pipelines runs rm is the
CLI form; retention apply now shares the complete row deletion.
Auto runs no longer alert on failure or success; only notify_mode on does,
unless the run_finished setting (now 'Notify when any run ends', default off)
opts into every run end. Waiting still notifies auto and on runs. The setting
is stored under a new key so a value saved under the old meaning is ignored.
When canonical main is not an ancestor of the item HEAD, the gate now
merges main into the item workspace under the delivery lock, re-runs the
declared checks on the merged commit, and fast-forwards main to it. A
conflicting merge is aborted with the workspace verified unchanged and a
diagnostic naming the files. Successful catch-ups are logged.
Later visits' input.md held only the previous stage's output, so the
operator's task reached only the first stage; when a Prepare stage
summarised (paths, HEAD), Code never saw the task or its acceptance
checklist and either blocked (run #202) or did part of it (#195, #197,
#198). Later inputs now carry the run prompt under '# Run task' (headings
demoted) before '# Previous turn output'.
Persist three global ntfy switches in server_settings: run_finished
(default on: auto-mode runs notify on success, read inside the
transition transaction; on/off modes unchanged), chat_needs_you
(default on: gates chat question/approval alerts) and
chat_reply_finished (default off: one alert per finished assistant
turn). Expose notification_settings / set_notification_settings over
the websocket protocol, add a Phone notifications section with
Material and Cupertino checkboxes in Settings, document the settings,
and keep the cli, tui, stage and pipeline test packages from
inheriting a real ntfy configuration.
Run notifications now always click through to pipelines://run/<id>; a valid
PIPELINES_WEB_PUBLIC_URL adds an "Open in browser" ntfy view action instead
of replacing the click. Queued rows derive links from the run id at delivery,
so legacy rows holding a web URL still open the app.
Unresolved chat questions/forms and permission requests publish one async,
bounded notification to the system destination (pipelines://chat/<id> plus the
browser action), deduplicated per form/request id across reconnects and
replays. stage.PublishNtfy gains optional, strictly validated view actions.
Server tests now clear the operator's real ntfy environment.