flowjet 2.0.0__tar.gz → 2.0.2__tar.gz

{flowjet-2.0.0 → flowjet-2.0.2}/CHANGELOG.md

@@ -8,6 +8,55 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 > History from 0.1.0–0.2.0 below was released as the separate `flowjet-server`
 > distribution, now merged into `flowjet`. See [docs/upgrade.md](docs/upgrade.md).
 
+## [Unreleased]
+
+## [2.0.2] — 2026-09-18
+
+### Added
+
+- **Duo backends — `soothe` or `soothe-nano`.** `FLOWJET_BACKEND` (CLI, default
+  `soothe`) and `FLOWJET_SERVER_BACKEND` (HTTP/ACP, default `nano`) select which
+  runtime builds the agent. `soothe` is a host layer over soothe-nano, so the
+  engine stays installed either way; the two surfaces switch independently
+  because the server needs a dual-mode agent (AGENT + ASK graphs per request),
+  which soothe does not expose — with `FLOWJET_SERVER_BACKEND=soothe` the server
+  builds one host agent per mode and routes on `interaction_mode`.
+
+### Changed
+
+- The CLI now builds its agent through the `soothe` host runtime by default
+  (host middleware, `request_plan_mode` / `ask_user` tools, configurable
+  persona). Set `FLOWJET_BACKEND=nano` for the previous behaviour.
+
+## [2.0.1] — 2026-09-17
+
+### Added
+
+- **ACP sessions run in the client's directory.** `FlowJetAcpAgent` now records
+  the `cwd` from `session/new`, `session/load`, `session/resume` and
+  `session/fork` and forwards it as the run's `metadata.workspace`. Previously
+  the client `cwd` was accepted and discarded, so every ACP session ran in a
+  hashed workspace under `FLOWJET_HOME`.
+
+### Changed
+
+- **The assistant identifies as FlowJet, not Soothe.** soothe-nano prepends its
+  own `<ASSISTANT_IDENTITY>` block — "a helpful AI assistant" named `Soothe` —
+  to every system prompt tier, and only parameterises it by `agent.name`.
+  `flowjet.core.identity` now installs a FlowJet template (coding agent, creator
+  line, and an explicit "never identify as Soothe / soothe-nano / Claude /
+  ChatGPT / Gemini / Anthropic / OpenAI / Google" rule) and `apply_profile`
+  defaults `agent.name` to `FlowJet` on both the CLI and server surfaces. An
+  explicit `agent.name` or `agent.system_prompt` in `nano.yml` still wins.
+- **Workspace resolution is caller-driven by default.**
+  `FLOWJET_ALLOW_EXTERNAL_WORKSPACE` now defaults to `true`, so a workspace
+  outside `FLOWJET_HOME` is honoured rather than rejected. This is not a
+  sandbox escape — tools remain confined to whichever workspace is selected
+  (`allow_paths_outside_workspace` is still `false`); the boundary moves to the
+  caller's directory instead of disappearing. Set
+  `FLOWJET_ALLOW_EXTERNAL_WORKSPACE=false` to restore forced hashed
+  per-session workspaces.
+
 ## [2.0.0] — 2026-09-17 — unified CLI + server
 
 **Breaking.** `flowjet-server` is merged into this repository and ships as one

flowjet-2.0.2/PKG-INFO

@@ -0,0 +1,195 @@
+Metadata-Version: 2.5
+Name: flowjet
+Version: 2.0.2
+Summary: FlowJet — agent CLI (fj) and server (ACP + OpenAI-compatible HTTP) powered by soothe-nano
+Project-URL: Homepage, https://github.com/caesar0301/flowjet
+Project-URL: Repository, https://github.com/caesar0301/flowjet
+Project-URL: Issues, https://github.com/caesar0301/flowjet/issues
+Author: Xiaming Chen
+License: MIT
+License-File: LICENSE
+Keywords: acp,agent,ai,cli,coding-agent,fj,flowjet,nano,openai,server,soothe
+Classifier: Development Status :: 4 - Beta
+Classifier: Environment :: Console
+Classifier: Environment :: Web Environment
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Requires-Python: <3.15,>=3.11
+Requires-Dist: soothe-nano<1.3.0,>=1.2.28
+Requires-Dist: soothe<2.0.0,>=1.0.12
+Provides-Extra: dev
+Requires-Dist: httpx>=0.27.0; extra == 'dev'
+Requires-Dist: mypy>=1.0.0; extra == 'dev'
+Requires-Dist: openai>=2.51.0; extra == 'dev'
+Requires-Dist: pytest-asyncio>=1.3.0; extra == 'dev'
+Requires-Dist: pytest-cov; extra == 'dev'
+Requires-Dist: pytest>=8.0.0; extra == 'dev'
+Requires-Dist: ruff>=0.12.0; extra == 'dev'
+Requires-Dist: twine>=5.0.0; extra == 'dev'
+Provides-Extra: server
+Requires-Dist: agent-client-protocol[http]<1.0.0,>=0.12.1; extra == 'server'
+Requires-Dist: fastapi>=0.115.0; extra == 'server'
+Requires-Dist: langchain-core>=0.3; extra == 'server'
+Requires-Dist: pydantic-settings>=2.0; extra == 'server'
+Requires-Dist: pydantic>=2.0; extra == 'server'
+Requires-Dist: uvicorn[standard]>=0.32.0; extra == 'server'
+Description-Content-Type: text/markdown
+
+# FlowJet
+
+[![PyPI version](https://img.shields.io/pypi/v/flowjet.svg)](https://pypi.org/p/flowjet)
+[![Python versions](https://img.shields.io/pypi/pyversions/flowjet.svg)](https://pypi.org/p/flowjet)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
+
+**FlowJet** is a coding agent you can run two ways:
+
+- **`fj`** — ask a question in your terminal, get an answer. No UI, no context-switching.
+- **`flowjet-server`** — the same agent as a service: **ACP** over WebSocket and an **OpenAI-compatible HTTP API**, so any OpenAI SDK can drive it.
+
+It runs on [soothe-nano](https://github.com/mirasoth/soothe-nano) — tools, skills, MCP, subagents — with SQLite persistence, so every conversation is resumable.
+
+---
+
+## Install
+
+```bash
+pip install flowjet              # CLI
+pip install 'flowjet[server]'    # CLI + HTTP service
+```
+
+Requires Python 3.11+. The server is an optional extra — installing just `flowjet` pulls a single runtime dependency.
+
+## Configure
+
+```bash
+fj setup                         # guided: pick a local or hosted model
+# or, with no config file at all:
+export OPENAI_API_KEY=sk-...
+fj summarize README.md
+```
+
+`fj setup` writes `~/.soothe/config/nano.yml`. Without it, FlowJet falls back to `OPENAI_API_KEY` or `ANTHROPIC_API_KEY`.
+
+Not sure your machine is ready? `fj doctor` (add `--deep`, `--live-llm`).
+
+### Backends
+
+Two runtimes can build the agent; they are switched independently per surface:
+
+| Env var | Surface | Default | Values |
+|---------|---------|---------|--------|
+| `FLOWJET_BACKEND` | CLI (`fj`) | `soothe` | `soothe`, `nano` |
+| `FLOWJET_SERVER_BACKEND` | `flowjet-server` / `fj serve` | `nano` | `soothe`, `nano` |
+
+`soothe` is the host runtime (extra middleware and tools over the same
+soothe-nano engine). The server defaults to `nano` because it needs a dual-mode
+agent — AGENT and ASK graphs selected per request — which `soothe` does not
+expose yet; on `soothe` the server builds one host agent per mode instead.
+
+---
+
+## CLI
+
+```bash
+fj explain this repo
+fj -f and now add tests          # continue this project's latest conversation
+fjf what did we decide?          # short alias of fj -f
+fj -l                            # list recent conversations
+```
+
+| Flag | Meaning |
+|------|---------|
+| `-f` / `--follow` | Continue the latest thread in this project |
+| `-t ID` / `--thread` | Continue (or pin) a specific thread — overrides `-f` |
+| `-l` / `--list` | List recent threads (newest first) |
+| `-n NUM` | How many threads `-l` shows (`0` = all) |
+| `-a` / `--ask` | Read-only: answer without touching files |
+| `-v` / `--verbose` | Mirror tool calls on stderr |
+| `-c PATH` / `-w DIR` | Alternate `nano.yml` / workspace root |
+
+`-f` is scoped to your current project, so work in other checkouts never hijacks your thread.
+
+Enable shell completion (predicts full queries, not just flags):
+
+```bash
+eval "$(fj completion zsh)"      # or: fj completion bash
+```
+
+---
+
+## Server
+
+```bash
+flowjet-server                   # host :: (dual-stack), port 8618
+fj serve --port 8618             # same, from the CLI
+```
+
+```python
+from openai import OpenAI
+
+client = OpenAI(api_key="local", base_url="http://127.0.0.1:8618/v1")
+print(client.responses.create(model="default", input="Hello"))
+```
+
+| Endpoint | Purpose |
+|----------|---------|
+| `POST /v1/responses` | Create a response (`stream=true` → SSE) |
+| `GET` / `DELETE /v1/responses/{id}` | Retrieve / cancel |
+| `POST /v1/chat/completions` | Chat Completions (non-stream + SSE) |
+| `GET /v1/models` | Model ids |
+| `GET /health` | Liveness |
+| `WS /acp` | Agent Client Protocol |
+
+Per-request options go in `extra_body.flowjet`:
+
+| Field | Values | Meaning |
+|-------|--------|---------|
+| `session` | string | Isolates workspace + conversation thread |
+| `interaction_mode` | `agent` (default) \| `ask` | `ask` is hard read-only |
+| `projection` | `report` \| `progress` \| `developer` | How much SSE detail to expose |
+
+Full configuration, environment variables and the security model: [docs/server.md](docs/server.md) · deployment: [deploy/README.md](deploy/README.md).
+
+---
+
+## Extend
+
+Add skills and MCP servers in `nano.yml`:
+
+```yaml
+skills:
+  - ~/.soothe/skills/my-reviewer
+
+mcp_servers:
+  - name: filesystem
+    transport: stdio
+    command: npx
+    args: ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
+```
+
+Skills are `SKILL.md` files loaded on demand; MCP tools activate on demand by default (`defer: true`).
+
+---
+
+## Development
+
+```bash
+git clone https://github.com/caesar0301/flowjet.git
+cd flowjet
+make sync-dev && make test && make lint
+```
+
+## Upgrading from 1.x
+
+Version 2.0 merges the former `flowjet-server` project into this one. The distribution is now `flowjet` and import paths changed. See [docs/upgrade.md](docs/upgrade.md).
+
+## License
+
+MIT

flowjet-2.0.2/README.md

@@ -0,0 +1,151 @@
+# FlowJet
+
+[![PyPI version](https://img.shields.io/pypi/v/flowjet.svg)](https://pypi.org/p/flowjet)
+[![Python versions](https://img.shields.io/pypi/pyversions/flowjet.svg)](https://pypi.org/p/flowjet)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
+
+**FlowJet** is a coding agent you can run two ways:
+
+- **`fj`** — ask a question in your terminal, get an answer. No UI, no context-switching.
+- **`flowjet-server`** — the same agent as a service: **ACP** over WebSocket and an **OpenAI-compatible HTTP API**, so any OpenAI SDK can drive it.
+
+It runs on [soothe-nano](https://github.com/mirasoth/soothe-nano) — tools, skills, MCP, subagents — with SQLite persistence, so every conversation is resumable.
+
+---
+
+## Install
+
+```bash
+pip install flowjet              # CLI
+pip install 'flowjet[server]'    # CLI + HTTP service
+```
+
+Requires Python 3.11+. The server is an optional extra — installing just `flowjet` pulls a single runtime dependency.
+
+## Configure
+
+```bash
+fj setup                         # guided: pick a local or hosted model
+# or, with no config file at all:
+export OPENAI_API_KEY=sk-...
+fj summarize README.md
+```
+
+`fj setup` writes `~/.soothe/config/nano.yml`. Without it, FlowJet falls back to `OPENAI_API_KEY` or `ANTHROPIC_API_KEY`.
+
+Not sure your machine is ready? `fj doctor` (add `--deep`, `--live-llm`).
+
+### Backends
+
+Two runtimes can build the agent; they are switched independently per surface:
+
+| Env var | Surface | Default | Values |
+|---------|---------|---------|--------|
+| `FLOWJET_BACKEND` | CLI (`fj`) | `soothe` | `soothe`, `nano` |
+| `FLOWJET_SERVER_BACKEND` | `flowjet-server` / `fj serve` | `nano` | `soothe`, `nano` |
+
+`soothe` is the host runtime (extra middleware and tools over the same
+soothe-nano engine). The server defaults to `nano` because it needs a dual-mode
+agent — AGENT and ASK graphs selected per request — which `soothe` does not
+expose yet; on `soothe` the server builds one host agent per mode instead.
+
+---
+
+## CLI
+
+```bash
+fj explain this repo
+fj -f and now add tests          # continue this project's latest conversation
+fjf what did we decide?          # short alias of fj -f
+fj -l                            # list recent conversations
+```
+
+| Flag | Meaning |
+|------|---------|
+| `-f` / `--follow` | Continue the latest thread in this project |
+| `-t ID` / `--thread` | Continue (or pin) a specific thread — overrides `-f` |
+| `-l` / `--list` | List recent threads (newest first) |
+| `-n NUM` | How many threads `-l` shows (`0` = all) |
+| `-a` / `--ask` | Read-only: answer without touching files |
+| `-v` / `--verbose` | Mirror tool calls on stderr |
+| `-c PATH` / `-w DIR` | Alternate `nano.yml` / workspace root |
+
+`-f` is scoped to your current project, so work in other checkouts never hijacks your thread.
+
+Enable shell completion (predicts full queries, not just flags):
+
+```bash
+eval "$(fj completion zsh)"      # or: fj completion bash
+```
+
+---
+
+## Server
+
+```bash
+flowjet-server                   # host :: (dual-stack), port 8618
+fj serve --port 8618             # same, from the CLI
+```
+
+```python
+from openai import OpenAI
+
+client = OpenAI(api_key="local", base_url="http://127.0.0.1:8618/v1")
+print(client.responses.create(model="default", input="Hello"))
+```
+
+| Endpoint | Purpose |
+|----------|---------|
+| `POST /v1/responses` | Create a response (`stream=true` → SSE) |
+| `GET` / `DELETE /v1/responses/{id}` | Retrieve / cancel |
+| `POST /v1/chat/completions` | Chat Completions (non-stream + SSE) |
+| `GET /v1/models` | Model ids |
+| `GET /health` | Liveness |
+| `WS /acp` | Agent Client Protocol |
+
+Per-request options go in `extra_body.flowjet`:
+
+| Field | Values | Meaning |
+|-------|--------|---------|
+| `session` | string | Isolates workspace + conversation thread |
+| `interaction_mode` | `agent` (default) \| `ask` | `ask` is hard read-only |
+| `projection` | `report` \| `progress` \| `developer` | How much SSE detail to expose |
+
+Full configuration, environment variables and the security model: [docs/server.md](docs/server.md) · deployment: [deploy/README.md](deploy/README.md).
+
+---
+
+## Extend
+
+Add skills and MCP servers in `nano.yml`:
+
+```yaml
+skills:
+  - ~/.soothe/skills/my-reviewer
+
+mcp_servers:
+  - name: filesystem
+    transport: stdio
+    command: npx
+    args: ["-y", "@modelcontextprotocol/server-filesystem", "/tmp"]
+```
+
+Skills are `SKILL.md` files loaded on demand; MCP tools activate on demand by default (`defer: true`).
+
+---
+
+## Development
+
+```bash
+git clone https://github.com/caesar0301/flowjet.git
+cd flowjet
+make sync-dev && make test && make lint
+```
+
+## Upgrading from 1.x
+
+Version 2.0 merges the former `flowjet-server` project into this one. The distribution is now `flowjet` and import paths changed. See [docs/upgrade.md](docs/upgrade.md).
+
+## License
+
+MIT

{flowjet-2.0.0 → flowjet-2.0.2}/docs/naming.md

@@ -10,7 +10,7 @@
 |--------|-------|------|
 | Product brand | **FlowJet** | Human-facing product name |
 | PyPI / `[project].name` | **`flowjet`** | Installable distribution (CLI **and** server) |
-| Source repository | **`flowjet-agent`** | GitHub repo (`caesar0301/flowjet-agent`) |
+| Source repository | **`flowjet`** | GitHub repo (`caesar0301/flowjet`) |
 | Formal CLI | **`flowjet`** | Canonical console entrypoint |
 | CLI alias | **`fj`** | Same as `flowjet` |
 | CLI alias | **`fjf`** | Same as `flowjet -f` / `fj -f` |
@@ -51,7 +51,7 @@ alias packages. Rationale:
 **Cost.** Anyone importing `fj_ai.*` or `flowjet_server.*` must update.
 Migration table: [upgrade.md](upgrade.md).
 
-**Not changed.** The repository name (`flowjet-agent`), the `fj` / `fjf` aliases,
+**Not changed.** The `fj` / `fjf` aliases,
 and the `flowjet-server` console script — so Docker, compose and existing shell
 habits keep working.
 

flowjet-2.0.2/docs/server.md

@@ -0,0 +1,136 @@
+# FlowJet server reference
+
+Everything you need to run `flowjet-server` (or `fj serve`) beyond the
+[quick start in the README](../README.md#server).
+
+## Endpoints
+
+Base path for the OpenAI-compatible API is `/v1`, so an SDK `base_url` should
+end in `/v1`.
+
+| HTTP | OpenAI SDK | Purpose |
+|------|-----------|---------|
+| `GET /health` | — | Liveness + thread-pool metrics |
+| `GET /v1/models` | `client.models.list()` | Logical model ids |
+| `POST /v1/responses` | `client.responses.create(...)` | Create a response |
+| `POST /v1/responses` + SSE | `client.responses.stream(...)` | Streaming response |
+| `GET /v1/responses/{id}` | `client.responses.retrieve(id)` | Retrieve a stored response |
+| `DELETE /v1/responses/{id}` | `client.responses.delete(id)` | Delete + cancel the run |
+| `POST /v1/chat/completions` | `client.chat.completions.create(...)` | Chat Completions (non-stream + SSE) |
+| `WS /acp` | — | Agent Client Protocol (JSON-RPC) |
+
+Responses are held in an in-process store; agent conversation state is
+persisted by soothe-nano (SQLite by default, Postgres with `deploy/nano.yml`).
+
+## `extra_body.flowjet`
+
+| Field | Values | Default | Meaning |
+|-------|--------|---------|---------|
+| `projection` | `report` \| `progress` \| `developer` | `report` | How much SSE detail to expose |
+| `session` | string | new `fj-<uuid>` | Isolates workspace + conversation thread |
+| `interaction_mode` | `agent` \| `ask` | `agent` | `ask` is hard read-only |
+| `metadata` | object | — | Opaque bag forwarded to the runtime |
+
+### Projection modes
+
+- **`report`** — final answer plus standard OpenAI lifecycle events.
+- **`progress`** — adds `response.flowjet.progress` stages.
+- **`developer`** — adds `response.flowjet.tool.started` / `.completed` summaries.
+
+No mode exposes chain-of-thought or tool arguments.
+
+### Ask vs agent
+
+The service builds nano's dual-mode agent and routes each request by
+`interaction_mode`:
+
+| Mode | Behavior |
+|------|----------|
+| `agent` | Full coding agent (read/write tools, shell if enabled in `nano.yml`) |
+| `ask` | Read-only: `ls`, `read_file`, `glob`, `grep`, `file_info` — no writes, no shell |
+
+Mode is pinned per session. To switch, start a new session.
+
+```python
+with client.responses.stream(
+    model="default",
+    input="Explain how auth works in this workspace.",
+    extra_body={"flowjet": {"projection": "progress", "interaction_mode": "ask"}},
+) as stream:
+    for event in stream:
+        if event.type == "response.flowjet.progress":
+            print(event.stage, event.message)
+        elif event.type == "response.output_text.delta":
+            print(event.delta, end="", flush=True)
+```
+
+## Environment
+
+| Variable | Default | Meaning |
+|----------|---------|---------|
+| `FLOWJET_API_KEY` | unset | If set, require Bearer auth |
+| `FLOWJET_MODELS` | `default` | Comma-separated logical model ids |
+| `FLOWJET_HOST` | `::` | Bind host (dual-stack IPv4/IPv6) |
+| `FLOWJET_PORT` | `8618` | Bind port |
+| `FLOWJET_HOME` | `~/.flowjet` | Root for per-session workspaces |
+| `FLOWJET_NANO_CONFIG` | unset | Optional `nano.yml` (else `$SOOTHE_HOME/config/nano.yml`) |
+| `FLOWJET_THREAD_POOL_MIN` | `2` | Min isolation worker threads |
+| `FLOWJET_THREAD_POOL_MAX` | `8` | Max isolation worker threads |
+| `FLOWJET_THREAD_POOL_IDLE_TIMEOUT` | `300` | Scaled-worker idle exit seconds (`0` = never) |
+| `FLOWJET_MAX_REQUESTS_PER_WORKER` | `100` | Recycle worker after N turns (`0` = unlimited) |
+| `FLOWJET_REUSE_RUNNER` | `true` | Reuse agent adapter per worker |
+| `FLOWJET_REQUEST_TIMEOUT` | `0` | Per-run timeout seconds (`0` = none — **set this in production**) |
+| `FLOWJET_READY_TIMEOUT` | `30` | Max wait for post-turn worker ready |
+| `FLOWJET_ALLOW_EXTERNAL_WORKSPACE` | `true` | Honour `metadata.workspace` outside `FLOWJET_HOME`; set `false` to force hashed per-session workspaces |
+| `FLOWJET_CORS_ORIGINS` | `*` | Comma-separated browser origins allowed to call the API |
+
+## Security model
+
+The CLI and the server have deliberately different invariants, declared in
+`flowjet.core.modes` rather than implied by whichever module builds the agent.
+
+| | CLI | Server |
+|---|---|---|
+| `allow_paths_outside_workspace` | `True` (your own machine) | **`False`** — per-request workspace, non-overridable |
+| `bypass` interaction mode | available | **rejected** — tools must never escape the request workspace |
+| Authentication | — | optional `FLOWJET_API_KEY` (Bearer, constant-time comparison) |
+
+Notes:
+
+- The server forces `allow_paths_outside_workspace = false` even when the
+  loaded `nano.yml` asks for wider access.
+- **The workspace itself is caller-chosen by default.** An ACP client's
+  `session/new` `cwd`, or `metadata.workspace` on an HTTP request, is honoured
+  even outside `FLOWJET_HOME` (`FLOWJET_ALLOW_EXTERNAL_WORKSPACE=true`). That
+  moves the confinement boundary to the caller's directory — it does not
+  remove it: tools still cannot escape the selected workspace. Set
+  `FLOWJET_ALLOW_EXTERNAL_WORKSPACE=false` to reject external paths and force
+  the hashed per-session workspace instead.
+- `FLOWJET_CORS_ORIGINS` defaults to `*` with credentials enabled so browser
+  clients work out of the box. Set an explicit origin list in production.
+- `/acp` (WebSocket) is unauthenticated unless `FLOWJET_API_KEY` is set;
+  put it behind a reverse proxy if you expose it publicly.
+
+## Concurrency
+
+Each session admits one in-flight request at a time; extra requests for the
+same session queue rather than interleave. Requests execute on a persistent
+thread pool (each worker owns an event loop), so long tool calls never block
+the HTTP loop. Tune with the `FLOWJET_THREAD_POOL_*` variables.
+
+The session's workspace is the caller's (`session/new` `cwd`, or
+`metadata.workspace`), falling back to `$FLOWJET_HOME/data/workspaces/ws_<hash>`
+when the caller does not name one.
+
+## Deploy
+
+Docker image and compose stack (Postgres + pgvector) are in
+[deploy/README.md](../deploy/README.md). The image installs with
+`--extra server` and keeps `CMD ["flowjet-server"]` plus a `:8618/health`
+healthcheck.
+
+## Specification
+
+- [RFC-001: OpenAI-Compatible API](specs/RFC-001-openai-compatible-api.md)
+- [RFC-002: Isolated Thread-Pool Runtime](specs/RFC-002-isolated-thread-pool-runtime.md)
+- [RFC-003: Production Isolation Hardening](specs/RFC-003-production-isolation-hardening.md)

{flowjet-2.0.0 → flowjet-2.0.2}/docs/specs/RFC-003-production-isolation-hardening.md

@@ -3,7 +3,7 @@
 **Status**: Draft  
 **Authors**: FlowJet  
 **Created**: 2026-07-31  
-**Last Updated**: 2026-07-31  
+**Last Updated**: 2026-09-17  
 **Depends on**: [RFC-002](RFC-002-isolated-thread-pool-runtime.md)  
 **Supersedes**: —  
 **Kind**: Architecture Design  
@@ -104,11 +104,17 @@ After each turn’s stream task finishes (success, cancel, error, timeout), canc
 
 ### 5.6 Workspace policy
 
-1. Default hashed path under `$FLOWJET_HOME/data/workspaces/` — unchanged.
+1. Default hashed path under `$FLOWJET_HOME/data/workspaces/` — used when the caller names no workspace.
 2. If `metadata.workspace` is set:
-   - If `FLOWJET_ALLOW_EXTERNAL_WORKSPACE=true` → resolve/create as today.
+   - If `FLOWJET_ALLOW_EXTERNAL_WORKSPACE=true` (**the default**) → resolve/create as today.
    - Else path MUST resolve under `$FLOWJET_HOME` (or equal home); otherwise reject with a clear runtime/OpenAI error (`invalid_workspace`).
 
+   Revised 2026-09-17: the flag's default flipped from `false` to `true` so a
+   desktop ACP client's `session/new` `cwd` is the session workspace without
+   requiring configuration. Tool confinement
+   (`allow_paths_outside_workspace=false`) is unchanged, so the boundary moves
+   to the caller's directory rather than being removed.
+
 ### 5.7 Interaction mode pin
 
 First admitted turn for a `session` records `interaction_mode` (`agent`|`ask`). Later turns with a different mode MUST fail with `interaction_mode_conflict` (HTTP 400). Omitted mode continues the pinned value (default `agent` if never set).
@@ -139,7 +145,7 @@ Expose a read-only snapshot: `total`, `idle`, `busy`, `dead`, `requests_complete
 | `FLOWJET_REUSE_RUNNER` | `true` | Reuse adapter per worker |
 | `FLOWJET_REQUEST_TIMEOUT` | `0` | Per-run timeout (s); production SHOULD set |
 | `FLOWJET_READY_TIMEOUT` | `30` | Ready-barrier timeout (s) |
-| `FLOWJET_ALLOW_EXTERNAL_WORKSPACE` | `false` | Allow workspace paths outside `FLOWJET_HOME` |
+| `FLOWJET_ALLOW_EXTERNAL_WORKSPACE` | `true` | Honour workspace paths outside `FLOWJET_HOME` (`false` = force hashed per-session workspaces) |
 
 ---
 

{flowjet-2.0.0 → flowjet-2.0.2}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "flowjet"
-version = "2.0.0"
+version = "2.0.2"
 description = "FlowJet — agent CLI (fj) and server (ACP + OpenAI-compatible HTTP) powered by soothe-nano"
 readme = "README.md"
 license = { text = "MIT" }
@@ -29,6 +29,9 @@ classifiers = [
 # lives in the `server` extra, and `flowjet.cli` never imports it at module
 # scope (see tests/unit/test_cli_no_server_deps.py).
 dependencies = [
+    # Default CLI runtime (FLOWJET_BACKEND=soothe). soothe is a host layer over
+    # soothe-nano, so nano stays pinned explicitly: soothe only caps it at <2.
+    "soothe>=1.0.12,<2.0.0",
     "soothe-nano>=1.2.28,<1.3.0",
 ]
 
@@ -74,9 +77,9 @@ fjf = "flowjet.cli.cli:main_follow"
 flowjet-server = "flowjet.server.__main__:main"
 
 [project.urls]
-Homepage = "https://github.com/caesar0301/flowjet-agent"
-Repository = "https://github.com/caesar0301/flowjet-agent"
-Issues = "https://github.com/caesar0301/flowjet-agent/issues"
+Homepage = "https://github.com/caesar0301/flowjet"
+Repository = "https://github.com/caesar0301/flowjet"
+Issues = "https://github.com/caesar0301/flowjet/issues"
 
 [tool.hatch.build.targets.wheel]
 packages = ["src/flowjet"]