feat: containerize DeepSQL Agent as fifth Compose service

Enterprise self-host no longer needs a host-side agent runtime. The new
`deepsql-agent` service (agent/Dockerfile) serves the Agent tab, AI
dashboards, Slack/CLI agent turns, and per-user profile provisioning on
the compose network (:8787 / :8788).

- Add agent/Dockerfile + docker-entrypoint.sh (DeepSQL-branded surface)
- Wire deepsql-agent into docker-compose.yml; nginx proxies /agent-api to it
- Default AGENT_WEBUI_URL / AGENT_PROVISIONER_URL to deepsql-agent
- install.sh generates AGENT_PROVISION_SECRET and builds the agent image
- Host setup-agent.sh remains for native/dev only (opt-in)
- Docs, smoke-test, and status scripts refer to DeepSQL Agent throughout

Co-authored-by: Venkat SF <venkatesh.sakamuri@stayflexi.com>

Author: cursoragent

.dockerignore

@@ -0,0 +1,19 @@
+# Shared by frontend (./Dockerfile) and deepsql-agent (agent/Dockerfile),
+# both of which use the repo root as build context.
+.git
+.github
+**/.venv
+**/venv
+**/node_modules
+**/__pycache__
+backend/target
+backend/.mvn/wrapper/maven-wrapper.jar
+*.log
+.env
+.env.*
+!.env.example
+dist
+coverage
+tmp
+.cursor
+.hermes

.env.example

@@ -128,12 +128,23 @@ DEEPSQL_EMBEDDING_MODEL=text-embedding-3-large
 # profile which defaults to false.
 EMBEDDING_FAIL_OPEN=false
 
-# ── Hermes agent (Agent tab + AI dashboards) ────────────────────────────────
-# Backend → Hermes for dashboards / Slack / CLI. Compose defaults this; only
-# override if Hermes listens elsewhere. Requires scripts/self-host/setup-agent.sh
-# (also run by install.sh unless DEEPSQL_SKIP_AGENT_SETUP=1).
-#AGENT_WEBUI_URL=http://host.docker.internal:8787
-#DEEPSQL_SKIP_AGENT_SETUP=0
+# ── DeepSQL Agent (Agent tab + AI dashboards) ───────────────────────────────
+# The fifth Compose service (`deepsql-agent`) serves the Agent tab, AI
+# dashboards, and Slack/CLI agent turns. install.sh auto-generates
+# AGENT_PROVISION_SECRET; Compose wires the rest. Override only for native
+# (non-Compose) development or a custom agent hostname.
+#
+# AGENT_WEBUI_URL          Agent HTTP API. Compose default: http://deepsql-agent:8787
+# AGENT_PROVISIONER_URL   Per-user profile provisioner. Compose default:
+#                         http://deepsql-agent:8788/provision
+# AGENT_PROVISION_SECRET  Shared secret between backend and agent (required).
+# DEEPSQL_AGENT_PORT / DEEPSQL_AGENT_PROVISIONER_PORT — host port mappings.
+#
+#AGENT_WEBUI_URL=http://deepsql-agent:8787
+#AGENT_PROVISIONER_URL=http://deepsql-agent:8788/provision
+AGENT_PROVISION_SECRET=change-me-agent-provision-secret
+#DEEPSQL_AGENT_PORT=8787
+#DEEPSQL_AGENT_PROVISIONER_PORT=8788
 #DEEPSQL_SMOKE_AGENT=1
 
 # ── Demo Data Seeding ───────────────────────────────────────────────────────
@@ -156,31 +167,15 @@ EMBEDDING_FAIL_OPEN=false
 # code reads any azure.openai.* property any more, so setting them changes nothing.
 
 # ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-# OPTIONAL — DeepSQL Agent runtime (chat TUI + the web Agent tab)
+# OPTIONAL — Native (non-Compose) DeepSQL Agent
 # ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
-# The agent surfaces — `deepsql agent` and the Agent tab — are served by a separate
-# runtime, not by the backend. There is no deepsql-agent container in this stack:
-# setup-agent.sh installs Hermes on the *host*, listening on :8787. That is why the
-# chat default reaches out of the container instead of across the Compose network.
-# Everything else in DeepSQL works without any of this.
-#
-# AGENT_WEBUI_URL         where the agent runtime serves its API. Default
-#                         http://host.docker.internal:8787 (AgentChatClient.java:58),
-#                         which is right under Compose — but a native backend run has no
-#                         such host, so there it must be set to http://127.0.0.1:8787.
-# AGENT_PROVISIONER_URL   per-user profile provisioning endpoint. Default
-#                         http://deepsql-agent:8788/provision (AgentBridgeService.java:65)
-#                         still names the container that does not exist here, so this one
-#                         must be set explicitly — under Compose and natively alike.
-# AGENT_PROVISION_SECRET  shared secret for the above. Unset, the backend logs
-#                         "agent.provision-secret is unset — skipping" and never creates
-#                         the u-<user> profile, so the agent has no identity to run as.
-#
-# Native runs: start the provisioner with `python3 scripts/local-agent-provisioner.py`.
-# See AGENTS.md for the full sequence, and agent/README.md for the runtime itself.
+# Only needed when you run the backend with `mvn spring-boot:run` instead of
+# Compose. Point at a local agent process and start the provisioner:
+#   python3 scripts/local-agent-provisioner.py
+# Or use the legacy host installer: ./scripts/self-host/setup-agent.sh
+# (set DEEPSQL_HOST_AGENT_SETUP=1). Prefer Compose for self-host / enterprise.
 #AGENT_WEBUI_URL=http://127.0.0.1:8787
 #AGENT_PROVISIONER_URL=http://127.0.0.1:8788/provision
-#AGENT_PROVISION_SECRET=
 
 # ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 # VECTOR STORE — choose one mode

CLAUDE.md

@@ -54,6 +54,17 @@ docker compose down             # Stop
 
 **Vault DB**: `jdbc:postgresql://localhost:5432/dba_agent` (postgres/postgres)
 
+### Self-host Compose (5 services)
+
+```bash
+./scripts/self-host/install.sh   # builds + starts everything
+docker compose ps                # postgres, valkey, backend, deepsql-agent, frontend
+```
+
+The **DeepSQL Agent** is the fifth container (`agent/Dockerfile`): Agent tab, AI
+dashboards, Slack/CLI agent turns, and per-user profile provisioning on :8787/:8788.
+No host-side agent install is required for Compose deployments.
+
 ## Architecture
 
 ```
@@ -90,7 +101,7 @@ src/                # Frontend (React)
 
 docs/               # Documentation
 mcp/                # DeepSQL Phase 1 MCP server (Node stdio wrapper around backend APIs)
-agent/              # DeepSQL Agent customization (persona, skills, skins; customized Hermes runtime)
+agent/              # DeepSQL Agent (persona, skills, skins, Dockerfile for the Compose service)
 ```
 
 ## MCP Server

README.md

@@ -174,16 +174,15 @@ is rejected. Changing width means migrating the column.
 ./scripts/self-host/install.sh
 ```
 
-The installer generates your JWT secret, the credential-vault encryption key and the vault
-DB password, prompts for the first admin account, builds both images, starts the stack, and
-verifies pgvector is live. Unless you set `DEEPSQL_SKIP_AGENT_SETUP=1`, it also runs
-[`scripts/self-host/setup-agent.sh`](scripts/self-host/setup-agent.sh) to install Hermes under
-`~/.hermes/`, wire DeepSQL MCP, and start the webui on `0.0.0.0:8787` (required for the
-**Agent** tab and AI dashboard generation).
+The installer generates your JWT secret, the credential-vault encryption key, the vault
+DB password, and the DeepSQL Agent provision secret; prompts for the first admin account;
+builds the backend, frontend, and DeepSQL Agent images; starts the stack; and verifies
+pgvector is live. The **Agent** tab and AI dashboard generation are served by the
+`deepsql-agent` Compose service — no host-side agent install is required.
 
 **The first build takes several minutes** — it compiles the Spring Boot backend with Maven
-inside the container and bundles the frontend with Vite. It has not hung. Later builds reuse
-the Docker layer cache.
+inside the container, bundles the frontend with Vite, and builds the DeepSQL Agent image.
+It has not hung. Later builds reuse the Docker layer cache.
 
 Then open **http://localhost:3000** and log in with the admin email and password you entered.
 

agent/Dockerfile

@@ -0,0 +1,90 @@
+# DeepSQL Agent — self-contained container for the Agent tab, AI dashboards,
+# Slack/CLI agent turns, and per-user profile provisioning.
+#
+# Built from this checkout. The runtime engine is an upstream dependency
+# installed at image-build time; the product surface (persona, skills, MCP,
+# branding) is owned by DeepSQL and is what operators interact with.
+
+FROM python:3.12-slim-bookworm
+
+ENV DEBIAN_FRONTEND=noninteractive \
+    PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1 \
+    # Product home — operators never need to know the upstream layout.
+    DEEPSQL_AGENT_HOME=/var/lib/deepsql-agent \
+    DEEPSQL_AGENT_ROOT=/opt/deepsql-agent \
+    PATH="/opt/deepsql-agent/runtime/venv/bin:/usr/local/bin:${PATH}"
+
+RUN apt-get update && apt-get install -y --no-install-recommends \
+      ca-certificates \
+      curl \
+      git \
+      nodejs \
+      npm \
+      build-essential \
+    && rm -rf /var/lib/apt/lists/*
+
+# Node 20+ is required by the DeepSQL MCP server. Debian bookworm ships
+# Node 18; replace with NodeSource 20.
+RUN curl -fsSL https://deb.nodesource.com/setup_20.x | bash - \
+    && apt-get install -y --no-install-recommends nodejs \
+    && rm -rf /var/lib/apt/lists/* \
+    && node --version && npm --version
+
+# Install uv for fast Python env management.
+RUN curl -fsSL https://astral.sh/uv/install.sh | sh \
+    && ln -sf /root/.local/bin/uv /usr/local/bin/uv
+
+WORKDIR /opt/deepsql-agent
+
+# Upstream runtime clones. Refs are overridable at build time; defaults track
+# what scripts/self-host/setup-agent.sh installs for host-based installs.
+ARG AGENT_RUNTIME_REPO=https://github.com/NousResearch/hermes-agent.git
+ARG AGENT_RUNTIME_REF=main
+ARG AGENT_API_REPO=https://github.com/nesquena/hermes-webui.git
+ARG AGENT_API_REF=main
+
+# Runtime engine (Python agent)
+RUN git clone --depth 1 --branch "${AGENT_RUNTIME_REF}" "${AGENT_RUNTIME_REPO}" runtime \
+    && cd runtime \
+    && UV_NO_CONFIG=1 uv sync \
+    && if [ -d .venv ] && [ ! -d venv ]; then ln -sfn .venv venv; fi \
+    && if [ -d venv ] && [ ! -d .venv ]; then ln -sfn venv .venv; fi \
+    # Pin MCP SDK below 2.0 — SDK 2.x renamed CallToolResult.isError → is_error
+    # and breaks every DeepSQL tool call until the runtime catches up.
+    && UV_NO_CONFIG=1 uv pip install --python venv/bin/python 'mcp>=1.0,<2' 'pyyaml>=6'
+
+# HTTP API surface the frontend / backend talk to (:8787)
+RUN git clone --depth 1 --branch "${AGENT_API_REF}" "${AGENT_API_REPO}" api \
+    && cd api \
+    && if [ -f requirements.txt ]; then \
+         /opt/deepsql-agent/runtime/venv/bin/pip install -r requirements.txt; \
+       fi
+
+# DeepSQL product surface — persona, skills, branding, MCP, provisioner.
+COPY agent/SOUL.md              /opt/deepsql-agent/SOUL.md
+COPY agent/skills               /opt/deepsql-agent/skills
+COPY agent/webui                /opt/deepsql-agent/webui-overlay
+COPY agent/skins                /opt/deepsql-agent/skins
+COPY agent/distribution.yaml    /opt/deepsql-agent/distribution.yaml
+COPY mcp                        /opt/deepsql-agent/mcp
+COPY scripts/local-agent-provisioner.py /opt/deepsql-agent/provisioner.py
+COPY agent/docker-entrypoint.sh /opt/deepsql-agent/docker-entrypoint.sh
+
+RUN chmod +x /opt/deepsql-agent/docker-entrypoint.sh \
+    && cd /opt/deepsql-agent/mcp && npm install --omit=dev --ignore-scripts \
+    && mkdir -p /var/lib/deepsql-agent/logs /var/lib/deepsql-agent/profiles \
+    # Apply DeepSQL Agent branding to the API UI (idempotent overlay).
+    && bash /opt/deepsql-agent/webui-overlay/apply-overlay.sh /opt/deepsql-agent/api || true
+
+# Expose the agent API (:8787) and the secret-gated provisioner (:8788).
+EXPOSE 8787 8788
+
+VOLUME ["/var/lib/deepsql-agent"]
+
+# Provisioner /health is unauthenticated and always 200 when the process is up.
+# The API on :8787 may require a session, so we don't use it for the healthcheck.
+HEALTHCHECK --interval=30s --timeout=5s --start-period=90s --retries=5 \
+  CMD curl -fsS http://127.0.0.1:8788/health >/dev/null || exit 1
+
+ENTRYPOINT ["/opt/deepsql-agent/docker-entrypoint.sh"]

agent/README.md

@@ -44,20 +44,31 @@ agent HTTP API (`/agent-api/*`), not the Hermes webui skin.
 These mirror workflows the in-house `AgentOrchestrator` performed, re-expressed as
 agent persona + skills over the DeepSQL MCP tools.
 
-## Install (local / self-host)
+## Install (self-host / enterprise)
 
-1. Install the upstream agent runtime (see [AGENTS.md](../AGENTS.md) Cursor Cloud notes
-   or the [Hermes install docs](https://hermes-agent.nousresearch.com/)).
-2. Apply DeepSQL customization (requires `AZURE_OPENAI_KEY` in the environment or
+**Preferred:** the `deepsql-agent` Compose service. `./scripts/self-host/install.sh`
+builds `agent/Dockerfile` and starts the Agent API (:8787) + profile provisioner
+(:8788) on the compose network. No host-side agent install is required.
+
+```bash
+./scripts/self-host/install.sh
+# or
+docker compose up -d --build deepsql-agent
+```
+
+### Native / local development (optional)
+
+1. Install the upstream agent runtime (see [AGENTS.md](../AGENTS.md) Cursor Cloud notes).
+2. Apply DeepSQL customization (requires `DEEPSQL_CHAT_API_KEY` in the environment or
    the repo `.env`):
 
 ```bash
 bash agent/install.sh
 ```
 
-It configures `~/.hermes/config.yaml` from this repo:
+It configures the agent home from this repo:
 
-- **model** — Azure OpenAI via its OpenAI-compatible `…/openai/v1` endpoint (key from env/.env, never committed)
+- **model** — via OpenAI-compatible endpoint (key from env/.env, never committed)
 - **mcp_servers.deepsql** — this repo’s `mcp/deepsql-phase1-server.js`
 - **skills.external_dirs** — this repo’s `agent/skills` (source of truth; must be a YAML list)
 - **approvals.mode: smart**, **SOUL.md** persona, and disables host-affecting toolsets
@@ -66,11 +77,15 @@ It configures `~/.hermes/config.yaml` from this repo:
 Verify:
 
 ```bash
-cd ~/.hermes/hermes-agent && uv run hermes mcp test deepsql   # → Connected, DeepSQL tools
+# Compose
+curl -fsS http://localhost:8788/health
+
+# Native (after setup-agent.sh)
+./scripts/self-host/setup-agent.sh
 ```
 
 The DeepSQL MCP server and the Spring backend remain the DBA brain; the agent consumes them.
-Optional upstream webui skin: see [`webui/`](webui/).
+Optional upstream UI skin: see [`webui/`](webui/).
 
 ### Approval UX (operator note)