nisse/docs/Development-Plan.md

Nisse — Development Plan

Status: agreed (PR #84); reworked in PR #92 after the architecture revision (PR #90). How to get from the code as it is to the architecture in docs/Architecture.md, in what order, and how each step is accepted. The architecture says what and why; this document says how and when, and is revised as work lands.

The priority is live testing. Work is ordered so the agent runs live against real resources as early and as often as possible, and what each run finds steers what comes next.

---

1. Where we are

A survey of main (2026-10-02) against the architecture. Done means it exists and matches; partial, something exists but needs changing; none, not started.

Only nisse-dev has real code; nisse-core and nisse-ops are placeholders.

Section	Status	Notes
L1 Work model	partial	Work items table with an active pointer (M1.1); no issue intake, queue, parking or blocked items
L2 Agent loop	partial	control.py, a fixed state machine with no-stranding, stop, abandon. Replaced by the new loop (Phase 2)
L3 Authority	partial	Revision binding and arrival stamping (gates.py); scope is still plan approval, not the issue and its conversation
L4 Classifier	none
L5 Conversation	partial	Matrix transport, mentions, lenient commands; no threads, no question handling, little explanation
L6 Context	partial	Context fitting, per-role views and orchestrator facts (M1.2); no scope or agent memory
L7 Canonical log	partial	Work item and actor on every event, fact (M1.1, M1.2); no classifier_verdict, knowledge_retrieved
L8 Durability	mostly done	WAL store, intents, boot reconciliation, PR adoption, comment redelivery; to move onto the new loop
L9 Models and roles	partial	LiteLLM, delegation, ask_expert, eval capture; roles are still planner/executor, no capability profiles
L10 Knowledge vault	none
L11 Injection	partial	Operator allowlists on every channel; no classifier
L12 Secrets, logging	partial	Structured state-change log; secrets in .env; no per-operation handout or redaction
L13 Execution	none	Tools run locally in the agent's process. The token leaks into the clone (#89)
D1–D2 Workflow	none	Plans posted to a Gitea issue and gated; no issue intake, labels, or documents in the repo
D3–D4 Workspace, git	partial	Idempotent publishing; no fresh workspace (#62), git add -A (#80)
D5 Review	partial	PR conversation comments only (#69), watch stall (#70)
D7–D8 Deployment, projects	none	Runs from a venv as a systemd unit on this box; one project in config
O, X	none	No container image yet
specs/	done for now	orchestrator.qnt (current loop), loop.qnt (the architecture, #91)

---

2. The environment today

• Supporting services (X5): Matrix, Gitea, LiteLLM, Vaultwarden, LLDAP with Authelia. The infrastructure repository exists in Gitea.
• Local inference: llama-swap on an RTX 4090 host (320 W cap). The benchmark (PR #55) chose qwen3.8-27b for local work.
• Models by role today: executor qwen3.8-27b, planner and helper claude-opus-5-5, expert claude-haiku-4-5. Phase 2 moves to driver and helpers.
• Bring-up runs on this development VM (scripts/bringup.sh), against test resources: the sandbox repository, the test Matrix room, a budget-capped LiteLLM key; configuration in /root/nisse-e2e/.
• A build VM for bring-up is provided by the operator by hand until Terraform (Phase 6); the agent reaches it over SSH with a key from /root/nisse-e2e/ssh/.

---

3. How we work

• One PR per milestone, each starting with a short implementation plan per CLAUDE.md. A milestone marked needs discussion starts with a design discussion first.
• Every change comes with tests, through scripts/check.sh.
• A live run follows each milestone that changes behaviour, not just each phase: a short bring-up run of what the milestone added. Findings are filed as issues and either fixed before moving on or deliberately deferred.
• The Quint models move with the loop: orchestrator.qnt while control.py runs, loop.qnt for the new loop, whose properties become tests as each piece lands (Phase 2).
• control.py gets fixes only, and only for what blocks a live run. New behaviour goes into the new loop.
• Bring-up runs on this box as root, against test resources only, with today's gates on until the classifier and trust dial are proven (Phase 3).
• This plan is revised as we go, in the same PR as the work that changes it.

---

4. Phases

Phase 0 — Housekeeping (done)

M0.1 Retire the old spec. Done. M0.2 A bring-up script. Done.

Phase 1 — Execution over SSH, and the agent in a container

M1.1 Work items in the state store. Done.

M1.2 Transcript views. Done. (#78 closes when a live run confirms it.)

M1.3 Execution over SSH. Pre-approved.

• execute(host, command, *, secrets, stdin) as the only way to run anything; file tools over SSH; the host bound by the harness from the project (L13).
• The workspace lives on the build VM: cloned there, with no token in its configuration; the push key handed to each push only (D4, #89).
• Every secret handed out is redacted from tool output (L12).
• Tests against a throwaway sshd container.
• Acceptance: the existing executor and delegation tests pass over SSH; a live run of one milestone on the operator's build VM.

M1.4 Projects (D8). Pre-approved.

• The project entry in configuration: repository, default branch, build VM, label defaults, secrets by name with their visibility, egress additions.
• Acceptance: the bring-up configuration is one project entry; a second project needs only a second entry.

M1.5 The agent as a container image. Pre-approved.

• A Dockerfile for nisse-dev, stateless with a volume for state; a Gitea Actions workflow that builds it and pushes to Gitea's registry on each release tag (X4).
• bringup.sh runs the image instead of a venv.
• Acceptance: a tagged release produces an image; a live run from the image.

Phase 1 acceptance: a live run, from the image, of a milestone and a review comment on the build VM.

Phase 2 — The new loop, as a thin slice first

Built alongside control.py, reusing the state store, log, views, publisher, pollers and recovery. The first milestone is the smallest end-to-end path; each later one widens it and is followed by a live run.

M2.1 The slice. Needs discussion (the loop's shape in code).

• An issue assigned to dev-agent → a work item (D1); a restatement in the item's thread; a go-ahead bound to its revision (L3, L5); one milestone through start_milestone / finish_milestone on a harness branch (D4); the PR; a review comment answered (D5); the merge; the item closed.
• Includes what carries over: a fresh workspace at item start and harness-owned branches (D3, D4, #62); publishing only the work (#80); review intake, inline comments and the watch stall (#69, #70); the review-fix check after merge.
• loop.qnt's properties for these paths become tests.
• Acceptance: a live run of that path, plus a crash mid-publish.

M2.2 The queue. Pre-approved. Several items: states, park-and-switch with branches pushed, next-item order, blocked items from Gitea dependencies, a closed PR handled (L1, D4). Live run with two issues and a dependency.

M2.3 Threads and labels. Pre-approved. Routing by thread, ambiguous commands asked about (L5); bug grace periods, discuss, design; questions and plain-language go-aheads (D1). Live run.

M2.4 Multi-milestone work and documents. Pre-approved. Build plans in the repository, ARCHITECTURE.md kept current, waiting for merges between milestones (D2, D4). Live run of a two-milestone feature.

M2.5 Conversations and new projects. Needs discussion. Work started in conversation, record issues carrying their item's ID, the new-project flow and its project entry (D1, D8, #29). Live run of a small new project.

M2.6 Roles and capability profiles. Pre-approved. driver, helper, expert, classifier (L9); escalation triggers; delegation and ask_expert on the new loop.

M2.7 Conversation. Pre-approved. Explanations, progress in prose, asking mid-work (L5, #42, #77).

M2.8 Switch over. Pre-approved, after acceptance. nisse-dev runs the new loop; control.py, its tests and orchestrator.qnt are removed.

Phase 3 — Authority

M3.1 Action classes and the fast path. Needs discussion. Harmless, contained and dangerous actions per agent (P3, D6), static fast-path rules.

M3.2 The classifier. Needs discussion. L4, its model, fail-closed verdicts, overrides as an evaluation set.

M3.3 The trust dial. Needs discussion. Per project (D8); agent merging behind protection it can't satisfy, verified by a test (D4, D5, #65).

Phase 4 — The environment

M4.1 Toolchain preflight, M4.2 recording installs, M4.3 environment repair. Pre-approved. All on the build VM (D3).

Phase 5 — Memory

M5.1 The vault. Pre-approved. The folder, per-agent mounts, control tools, atomic writes and provenance (L10).

M5.2 Agent memory (needs discussion) and M5.3 scope memory (pre-approved): when notes are written and read, verbatim corrections (L6).

M5.4 The operator's copy. Pre-approved. Syncthing, directions per folder.

Phase 6 — Production

M6.1 The agent host. Needs discussion. Containers, volumes, snapshots, per-agent networks (X1, X2, X4).

M6.2 Build VMs by Terraform. Needs discussion. The template, VLAN and egress (D7, X2), collection when idle.

M6.3 Secrets and identity. Pre-approved. Per-agent secret sets on the host, LLDAP principal (L12).

M6.4 Delivery. Pre-approved. Watchtower, restarts gated on an idle flag (X4).

M6.5 The browser tool. Pre-approved. On the internal test network (D7).

Phase 7 — nisse-core and nisse-ops

M7.1 Extract nisse-core. M7.2 MCP servers (#37). M7.3 Approved step sequences (O2). M7.4 Ops intake and duties from the infrastructure repository (O5). M7.5 The ops tool VM and deployment (O6). M7.6 Requests between agents: nisse-dev asking for build VMs (L1, O5). All needs discussion except M7.1.

Not yet scheduled

From the architecture's open questions and older issues: filing issues from conversation; stacked PRs; releases; nisse-ops issuing credentials; a server database; recovering the agent host (an ops issue); more than one active work item; vision; one local model for every agent; several forges (#26); a room per project (#28).

---

5. Open issues and where they land

Issue	Lands in
#26 multiple forges	Not yet scheduled
#27 per-project state, clone on use	M1.3 (clone on the build VM), M1.4 (projects)
#28 room per project	Not yet scheduled (threads, M2.3, may make it unnecessary)
#29 create a project from scratch	M2.5 (ownership settled by D4, D8)
#37 MCP servers	M7.2
#42 per-turn commentary	M2.7
#62 branch at execution start	M2.1
#65 agent merges approved PRs	M3.3
#69 inline review comments	M2.1
#70 watch stall	M2.1
#77 conversational agent	M2.3 (questions), M2.7
#78 transcript scoping and attribution	Done in M1.1, M1.2; closes after a live run
#80 build artifacts in PRs	M2.1
#89 token in the clone's .git/config	M1.3

6. Implementation notes

Mechanism detail moved out of docs/Architecture.md in PR #90, which keeps only decisions and their reasons. Keyed by the architecture section each note implements; a milestone touching that section uses the note, and changes it here if the implementation differs.

• L1 switching: the grace period before switching away from an item waiting on the operator defaults to 10 minutes. When the driver chooses the next ready item and the choice fails or is invalid, fall back to the oldest.
• L3 revision stamping: in Matrix, a reply is stamped with the revision whose message had been sent when it arrived; on a Gitea issue, with the latest proposal comment older than the reply.
• L5 transport: the Transport interface is send, ask and receive. Intake backoff is exponential, capped at a few minutes.
• L6 fitting order: first shorten old tool results to their first and last lines, then drop the oldest events with a note saying so; never the system prompt, the current plan or the most recent turns. View labels: [orchestrator], [operator, Matrix], [operator, comment on PR #N], [helper report], [executor] narration.
• L7 rendering: content is stored as bytes or local paths and uploaded or inlined at render time, never as provider handles. Token budgeting and prompt-cache breakpoints are render-time concerns, never stored. Compaction summaries are first-class events so a model change never inherits opaquely compacted history. A reasoning "session" is a delegation, including one resumed after a crash.
• L8 state store: SQLite in WAL mode, each change committed and fsynced before the loop continues. A delegation's cut is clean when its last event is a completed result, its branch is at the expected commit, its caps are unbreached and external truth matches the log. "Anything else" offers continue, re-plan or abandon.
• L9 local models: request tool_choice only of models whose server declares it can enforce it; otherwise the loop enforces the contract (an explicit finish, prose treated as a violation). Briefings use the capability profile, never a model name or "weaker model".
• L10 vault: atomic writes are write-then-rename. Obsidian's settings folder (and .git, if history is kept) are not synced. A failed write is retried, or recorded and reconciled later.
• L12: secrets reach a command by environment or stdin. Headless agents get an Authelia single-factor exception scoped by source address, group and target. Logs are shipped to the host journal, not kept in containers.
• D3 preflight: a project's needs are read from its files (Cargo.toml, pyproject.toml, package.json) or a declaration in the repository.
• D4 publishing: standard excludes are configured in each clone, and only what the work produced is committed. The push credential is handed to each push only: a deploy key over stdin to a short-lived ssh-agent on the build VM, or a push-only account's HTTPS token as that command's secrets, read by a credential helper that names environment variables, never values.
• D7: test services use ports 18000–18099, so opening access later is a firewall change. The inference host is behind LiteLLM, so it is not a second egress exception.
• X4: the agent host's state dataset is ZFS, snapshotted on a schedule and on each approval. Tool VM templates use Terraform's bpg/proxmox provider with cloud-init. Containers restart unless stopped.
• X5: the local inference host is llama-swap on an RTX 4090 machine.