diff --git a/docker-compose.yml b/docker-compose.yml index bca9ab9..6206b2c 100644 --- a/docker-compose.yml +++ b/docker-compose.yml @@ -243,10 +243,7 @@ services: - EDGE_TUNNEL_FQDN=${EDGE_TUNNEL_FQDN:-} - EDGE_TUNNEL_FQDN_CHAT=${EDGE_TUNNEL_FQDN_CHAT:-} - EDGE_ROUTING_MODE=${EDGE_ROUTING_MODE:-subpath} - # Cost caps / rate limiting (#711) - - CHAT_MAX_REQUESTS_PER_HOUR=${CHAT_MAX_REQUESTS_PER_HOUR:-60} - - CHAT_MAX_REQUESTS_PER_DAY=${CHAT_MAX_REQUESTS_PER_DAY:-500} - - CHAT_MAX_TOKENS_PER_DAY=${CHAT_MAX_TOKENS_PER_DAY:-1000000} + # Rate limiting removed (#1084) ports: - "80:80" - "443:443" diff --git a/docker/chat/server.py b/docker/chat/server.py index b5252a7..6472a1d 100644 --- a/docker/chat/server.py +++ b/docker/chat/server.py @@ -23,7 +23,6 @@ The claude binary is expected to be mounted from the host at /usr/local/bin/clau """ import asyncio -import datetime import json import os import re @@ -61,10 +60,6 @@ EDGE_ROUTING_MODE = os.environ.get("EDGE_ROUTING_MODE", "subpath") # (acceptable during local dev; production MUST set this). FORWARD_AUTH_SECRET = os.environ.get("FORWARD_AUTH_SECRET", "") -# Rate limiting / cost caps (#711) -CHAT_MAX_REQUESTS_PER_HOUR = int(os.environ.get("CHAT_MAX_REQUESTS_PER_HOUR", 60)) -CHAT_MAX_REQUESTS_PER_DAY = int(os.environ.get("CHAT_MAX_REQUESTS_PER_DAY", 500)) -CHAT_MAX_TOKENS_PER_DAY = int(os.environ.get("CHAT_MAX_TOKENS_PER_DAY", 1000000)) # Allowed users - disinto-admin always allowed; CSV allowlist extends it _allowed_csv = os.environ.get("DISINTO_CHAT_ALLOWED_USERS", "") @@ -90,11 +85,6 @@ _sessions = {} # Pending OAuth state tokens: state -> expires (float) _oauth_states = {} -# Per-user rate limiting state (#711) -# user -> list of request timestamps (for sliding-window hourly/daily caps) -_request_log = {} -# user -> {"tokens": int, "date": "YYYY-MM-DD"} -_daily_tokens = {} # WebSocket message queues per user # user -> asyncio.Queue (for streaming messages to connected clients) @@ -213,69 +203,9 @@ def _fetch_user(access_token): return None -# ============================================================================= -# Rate Limiting Functions (#711) -# ============================================================================= - -def _check_rate_limit(user): - """Check per-user rate limits. Returns (allowed, retry_after, reason) (#711). - - Checks hourly request cap, daily request cap, and daily token cap. - """ - now = time.time() - one_hour_ago = now - 3600 - today = datetime.date.today().isoformat() - - # Prune old entries from request log - timestamps = _request_log.get(user, []) - timestamps = [t for t in timestamps if t > now - 86400] - _request_log[user] = timestamps - - # Hourly request cap - hourly = [t for t in timestamps if t > one_hour_ago] - if len(hourly) >= CHAT_MAX_REQUESTS_PER_HOUR: - oldest_in_window = min(hourly) - retry_after = int(oldest_in_window + 3600 - now) + 1 - return False, max(retry_after, 1), "hourly request limit" - - # Daily request cap - start_of_day = time.mktime(datetime.date.today().timetuple()) - daily = [t for t in timestamps if t >= start_of_day] - if len(daily) >= CHAT_MAX_REQUESTS_PER_DAY: - next_day = start_of_day + 86400 - retry_after = int(next_day - now) + 1 - return False, max(retry_after, 1), "daily request limit" - - # Daily token cap - token_info = _daily_tokens.get(user, {"tokens": 0, "date": today}) - if token_info["date"] != today: - token_info = {"tokens": 0, "date": today} - _daily_tokens[user] = token_info - if token_info["tokens"] >= CHAT_MAX_TOKENS_PER_DAY: - next_day = start_of_day + 86400 - retry_after = int(next_day - now) + 1 - return False, max(retry_after, 1), "daily token limit" - - return True, 0, "" - - -def _record_request(user): - """Record a request timestamp for the user (#711).""" - _request_log.setdefault(user, []).append(time.time()) - - -def _record_tokens(user, tokens): - """Record token usage for the user (#711).""" - today = datetime.date.today().isoformat() - token_info = _daily_tokens.get(user, {"tokens": 0, "date": today}) - if token_info["date"] != today: - token_info = {"tokens": 0, "date": today} - token_info["tokens"] += tokens - _daily_tokens[user] = token_info - def _parse_stream_json(output): - """Parse stream-json output from claude --print (#711). + """Parse stream-json output from claude --print. Returns (text_content, total_tokens). Falls back gracefully if the usage event is absent or malformed. @@ -1063,34 +993,13 @@ class ChatHandler(BaseHTTPRequestHandler): except IOError as e: self.send_error_page(500, f"Error reading file: {e}") - def _send_rate_limit_response(self, retry_after, reason): - """Send a 429 response with Retry-After header and HTMX fragment (#711).""" - body = ( - f'
' - f"Rate limit exceeded: {reason}. " - f"Please try again in {retry_after} seconds." - f"
" - ) - self.send_response(429) - self.send_header("Retry-After", str(retry_after)) - self.send_header("Content-Type", "text/html; charset=utf-8") - self.send_header("Content-Length", str(len(body.encode("utf-8")))) - self.end_headers() - self.wfile.write(body.encode("utf-8")) - + def handle_chat(self, user): """ Handle chat requests by spawning `claude --print` with the user message. - Enforces per-user rate limits and tracks token usage (#711). Streams tokens over WebSocket if connected. """ - # Check rate limits before processing (#711) - allowed, retry_after, reason = _check_rate_limit(user) - if not allowed: - self._send_rate_limit_response(retry_after, reason) - return - # Read request body content_length = int(self.headers.get("Content-Length", 0)) if content_length == 0: @@ -1127,9 +1036,6 @@ class ChatHandler(BaseHTTPRequestHandler): if not conv_id or not _validate_conversation_id(conv_id): conv_id = _generate_conversation_id() - # Record request for rate limiting (#711) - _record_request(user) - try: # Save user message to history _write_message(user, conv_id, "user", message) @@ -1194,14 +1100,6 @@ class ChatHandler(BaseHTTPRequestHandler): # Combine response parts response = "".join(response_parts) - # Track token usage - does not block *this* request (#711) - if total_tokens > 0: - _record_tokens(user, total_tokens) - print( - f"Token usage: user={user} tokens={total_tokens}", - file=sys.stderr, - ) - # Fall back to raw output if stream-json parsing yielded no text if not response: response = proc.stdout.getvalue() if hasattr(proc.stdout, 'getvalue') else "" @@ -1294,18 +1192,6 @@ class ChatHandler(BaseHTTPRequestHandler): self.send_error_page(401, "Unauthorized: no valid session") return - # Check rate limits before allowing WebSocket connection - allowed, retry_after, reason = _check_rate_limit(user) - if not allowed: - self.send_error_page( - 429, - f"Rate limit exceeded: {reason}. Retry after {retry_after}s", - ) - return - - # Record request for rate limiting - _record_request(user) - # Create message queue for this user _websocket_queues[user] = asyncio.Queue() @@ -1421,12 +1307,6 @@ def main(): print("forward_auth secret configured (#709)", file=sys.stderr) else: print("WARNING: FORWARD_AUTH_SECRET not set - verify endpoint unrestricted", file=sys.stderr) - print( - f"Rate limits (#711): {CHAT_MAX_REQUESTS_PER_HOUR}/hr, " - f"{CHAT_MAX_REQUESTS_PER_DAY}/day, " - f"{CHAT_MAX_TOKENS_PER_DAY} tokens/day", - file=sys.stderr, - ) httpd.serve_forever() diff --git a/lib/AGENTS.md b/lib/AGENTS.md index ae56bbe..a3fce35 100644 --- a/lib/AGENTS.md +++ b/lib/AGENTS.md @@ -30,7 +30,7 @@ sourced as needed. | `lib/git-creds.sh` | Shared git credential helper configuration. `configure_git_creds([HOME_DIR] [RUN_AS_CMD])` — writes a static credential helper script and configures git globally to use password-based HTTP auth (Forgejo 11.x rejects API tokens for `git push`, #361). **Retry on cold boot (#741)**: resolves bot username from `FORGE_TOKEN` with 5 retries (exponential backoff 1-5s); fails loudly and returns 1 if Forgejo is unreachable — never falls back to a wrong hardcoded default (exports `BOT_USER` on success). `repair_baked_cred_urls([--as RUN_AS_CMD] DIR ...)` — rewrites any git remote URLs that have credentials baked in to use clean URLs instead; uses `safe.directory` bypass for root-owned repos (#671). Requires `FORGE_PASS`, `FORGE_URL`, `FORGE_TOKEN`. | entrypoints (agents, edge) | | `lib/ops-setup.sh` | `setup_ops_repo()` — creates ops repo on Forgejo if it doesn't exist, configures bot collaborators, clones/initializes ops repo locally, seeds directory structure (vault, knowledge, evidence, sprints). Evidence subdirectories seeded: engagement/, red-team/, holdout/, evolution/, user-test/. Also seeds sprints/ for architect output. Exports `_ACTUAL_OPS_SLUG`. `migrate_ops_repo(ops_root, [primary_branch])` — idempotent migration helper that seeds missing directories and .gitkeep files on existing ops repos (pre-#407 deployments). | bin/disinto (init) | | `lib/ci-setup.sh` | `_install_cron_impl()` — installs crontab entries for bare-metal deployments (compose mode uses polling loop instead). `_create_forgejo_oauth_app()` — generic helper to create an OAuth2 app on Forgejo (shared by Woodpecker and chat). `_create_woodpecker_oauth_impl()` — creates Woodpecker OAuth2 app (thin wrapper). `_create_chat_oauth_impl()` — creates disinto-chat OAuth2 app, writes `CHAT_OAUTH_CLIENT_ID`/`CHAT_OAUTH_CLIENT_SECRET` to `.env` (#708). `_generate_woodpecker_token_impl()` — auto-generates WOODPECKER_TOKEN via OAuth2 flow. `_activate_woodpecker_repo_impl()` — activates repo in Woodpecker. All gated by `_load_ci_context()` which validates required env vars. | bin/disinto (init) | -| `lib/generators.sh` | Template generation for `disinto init`: `generate_compose()` — docker-compose.yml (**duplicate service detection**: tracks service names during generation, aborts with `ERROR: Duplicate service name '$name' detected` on conflict; detection state is reset between calls so idempotent reinvocation is safe, #850) (uses `codeberg.org/forgejo/forgejo:11.0` tag; `CLAUDE_BIN_DIR` volume mount removed from agents/llama services — only `reproduce` and `edge` still use the host-mounted CLI (#992); adds `security_opt: [apparmor:unconfined]` to all services for rootless container compatibility; Forgejo includes a healthcheck so dependent services use `condition: service_healthy` — fixes cold-start races, #665; adds `chat` service block with isolated `chat-config` named volume and `CHAT_HISTORY_DIR` bind-mount for per-user NDJSON history persistence (#710); injects `FORWARD_AUTH_SECRET` for Caddy↔chat defense-in-depth auth (#709); cost-cap env vars `CHAT_MAX_REQUESTS_PER_HOUR`, `CHAT_MAX_REQUESTS_PER_DAY`, `CHAT_MAX_TOKENS_PER_DAY` (#711); subdomain fallback comment for `EDGE_TUNNEL_FQDN_*` vars (#713); all `depends_on` now use `condition: service_healthy/started` instead of bare service names; all services now include `restart: unless-stopped` including the edge service — #768; agents service now uses `image: ghcr.io/disinto/agents:${DISINTO_IMAGE_TAG:-latest}` instead of `build:` (#429); `WOODPECKER_PLUGINS_PRIVILEGED` env var added to woodpecker service (#779); agents-llama conditional block gated on `ENABLE_LLAMA_AGENT=1` (#769); `agents-llama-all` compose service (profile `agents-llama-all`, all 7 roles: review,dev,gardener,architect,planner,predictor,supervisor) added by #801; agents service gains volume mounts for `./projects`, `./.env`, `./state`), `generate_caddyfile()` — Caddyfile (routes: `/forge/*` → forgejo:3000, `/woodpecker/*` → woodpecker:8000, `/staging/*` → staging:80; `/chat/login` and `/chat/oauth/callback` bypass `forward_auth` so unauthenticated users can reach the OAuth flow; `/chat/*` gated by `forward_auth` on `chat:8080/chat/auth/verify` which stamps `X-Forwarded-User` (#709); root `/` redirects to `/forge/`), `generate_staging_index()` — staging index, `generate_deploy_pipelines()` — Woodpecker deployment pipeline configs. Requires `FACTORY_ROOT`, `PROJECT_NAME`, `PRIMARY_BRANCH`. | bin/disinto (init) | +| `lib/generators.sh` | Template generation for `disinto init`: `generate_compose()` — docker-compose.yml (**duplicate service detection**: tracks service names during generation, aborts with `ERROR: Duplicate service name '$name' detected` on conflict; detection state is reset between calls so idempotent reinvocation is safe, #850) (uses `codeberg.org/forgejo/forgejo:11.0` tag; `CLAUDE_BIN_DIR` volume mount removed from agents/llama services — only `reproduce` and `edge` still use the host-mounted CLI (#992); adds `security_opt: [apparmor:unconfined]` to all services for rootless container compatibility; Forgejo includes a healthcheck so dependent services use `condition: service_healthy` — fixes cold-start races, #665; adds `chat` service block with isolated `chat-config` named volume and `CHAT_HISTORY_DIR` bind-mount for per-user NDJSON history persistence (#710); injects `FORWARD_AUTH_SECRET` for Caddy↔chat defense-in-depth auth (#709); subdomain fallback comment for `EDGE_TUNNEL_FQDN_*` vars (#713); all `depends_on` now use `condition: service_healthy/started` instead of bare service names; all services now include `restart: unless-stopped` including the edge service — #768; agents service now uses `image: ghcr.io/disinto/agents:${DISINTO_IMAGE_TAG:-latest}` instead of `build:` (#429); `WOODPECKER_PLUGINS_PRIVILEGED` env var added to woodpecker service (#779); agents-llama conditional block gated on `ENABLE_LLAMA_AGENT=1` (#769); `agents-llama-all` compose service (profile `agents-llama-all`, all 7 roles: review,dev,gardener,architect,planner,predictor,supervisor) added by #801; agents service gains volume mounts for `./projects`, `./.env`, `./state`), `generate_caddyfile()` — Caddyfile (routes: `/forge/*` → forgejo:3000, `/woodpecker/*` → woodpecker:8000, `/staging/*` → staging:80; `/chat/login` and `/chat/oauth/callback` bypass `forward_auth` so unauthenticated users can reach the OAuth flow; `/chat/*` gated by `forward_auth` on `chat:8080/chat/auth/verify` which stamps `X-Forwarded-User` (#709); root `/` redirects to `/forge/`), `generate_staging_index()` — staging index, `generate_deploy_pipelines()` — Woodpecker deployment pipeline configs. Requires `FACTORY_ROOT`, `PROJECT_NAME`, `PRIMARY_BRANCH`. | bin/disinto (init) | | `lib/backup.sh` | Factory backup creation. `backup_create ` — exports factory state: fetches all issues (open+closed) from the project and ops repos via Forgejo API, bundles the ops repo as a git bundle, and writes a tarball. Requires `FORGE_URL`, `FORGE_TOKEN`, `FORGE_REPO`, `FORGE_OPS_REPO`, `OPS_REPO_ROOT`. Sourced by `bin/disinto backup create` (#1057). | bin/disinto (backup create) | | `lib/disinto/backup.sh` | Factory backup restore. `backup_import ` — restores from a backup tarball: creates missing repos via Forgejo API, imports issues (idempotent — skips by number if present), unpacks ops repo git bundle. Idempotent: running twice produces same end state with no errors. Requires `FORGE_URL`, `FORGE_TOKEN`. Sourced by `bin/disinto backup import` (#1058). | bin/disinto (backup import) | | `lib/sprint-filer.sh` | Post-merge sub-issue filer for sprint PRs. Invoked by the `.woodpecker/ops-filer.yml` pipeline after a sprint PR merges to ops repo `main`. Parses ` ... ` blocks from sprint PR bodies to extract sub-issue definitions, creates them on the project repo using `FORGE_FILER_TOKEN` (narrow-scope `filer-bot` identity with `issues:write` only), adds `in-progress` label to the parent vision issue, and handles vision lifecycle closure when all sub-issues are closed. Uses `filer_api_all()` for paginated fetches. Idempotent: uses `` markers to skip already-filed issues. Requires `FORGE_FILER_TOKEN`, `FORGE_API`, `FORGE_API_BASE`, `FORGE_OPS_REPO`. | `.woodpecker/ops-filer.yml` (CI pipeline on ops repo) | diff --git a/lib/generators.sh b/lib/generators.sh index 581de8b..41e7e67 100644 --- a/lib/generators.sh +++ b/lib/generators.sh @@ -681,6 +681,55 @@ COMPOSEEOF - disinto-net command: ["echo", "staging slot — replace with project image"] + # Chat container — Claude chat UI backend (#705) + # Internal service only; edge proxy routes to chat:8080 + # Sandbox hardened per #706 — no docker.sock, read-only rootfs, minimal caps + # Rate limiting removed (#1084) + chat: + build: + context: ./docker/chat + dockerfile: Dockerfile + container_name: disinto-chat + restart: unless-stopped + read_only: true + tmpfs: + - /tmp:size=64m + security_opt: + - no-new-privileges:true + cap_drop: + - ALL + pids_limit: 128 + mem_limit: 512m + memswap_limit: 512m + volumes: + # Mount claude binary from host (same as agents) + - ${CLAUDE_BIN_DIR}:/usr/local/bin/claude:ro + # Throwaway named volume for chat config (isolated from host ~/.claude) + - chat-config:/var/chat/config + # Chat history persistence: per-user NDJSON files on bind-mounted host volume + - ${CHAT_HISTORY_DIR:-./state/chat-history}:/var/lib/chat/history + environment: + CHAT_HOST: "0.0.0.0" + CHAT_PORT: "8080" + FORGE_URL: http://forgejo:3000 + CHAT_OAUTH_CLIENT_ID: ${CHAT_OAUTH_CLIENT_ID:-} + CHAT_OAUTH_CLIENT_SECRET: ${CHAT_OAUTH_CLIENT_SECRET:-} + EDGE_TUNNEL_FQDN: ${EDGE_TUNNEL_FQDN:-} + EDGE_TUNNEL_FQDN_CHAT: ${EDGE_TUNNEL_FQDN_CHAT:-} + EDGE_ROUTING_MODE: ${EDGE_ROUTING_MODE:-subpath} + DISINTO_CHAT_ALLOWED_USERS: ${DISINTO_CHAT_ALLOWED_USERS:-} + # Shared secret for Caddy forward_auth verify endpoint (#709) + FORWARD_AUTH_SECRET: ${FORWARD_AUTH_SECRET:-} + # Rate limiting removed (#1084) + healthcheck: + test: ["CMD", "python3", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8080/health')"] + interval: 30s + timeout: 5s + retries: 3 + start_period: 10s + networks: + - disinto-net + volumes: forgejo-data: woodpecker-data: diff --git a/nomad/jobs/chat.hcl b/nomad/jobs/chat.hcl index ad18cec..95f86ab 100644 --- a/nomad/jobs/chat.hcl +++ b/nomad/jobs/chat.hcl @@ -119,9 +119,7 @@ job "chat" { # CHAT_OAUTH_CLIENT_ID, CHAT_OAUTH_CLIENT_SECRET, FORWARD_AUTH_SECRET # rendered from kv/disinto/shared/chat via template stanza. env { - FORGE_URL = "http://forgejo:3000" - CHAT_MAX_REQUESTS_PER_HOUR = "60" - CHAT_MAX_REQUESTS_PER_DAY = "1000" + FORGE_URL = "http://forgejo:3000" } # ── Vault-templated secrets (S5.2, issue #989) ─────────────────────────