Compare commits

...

3 Commits

Author SHA1 Message Date
didericis 2cdedbb7ca docs(prd): add PRD for egress control plane
lint / lint (push) Successful in 2m13s
Out-of-band egress enforcement & cost-control plane: meter token usage
at the egress proxy, evaluate budgets with agent→bottle→parent→global
precedence, and force cutoff/freeze/kill without the agent in the loop.
Introduces a host-level SQLite ledger behind a thin repository API and a
host-only TUI dashboard. Closes the design discussion on #251.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NkwFXLFff9PYPy4wgVBJp9
2026-06-29 11:01:47 -04:00
didericis f787764364 refactor(manifest): break import cycle by extracting ManifestBottle to a leaf module
test / unit (pull_request) Successful in 57s
test / integration (pull_request) Successful in 27s
test / coverage (pull_request) Successful in 1m23s
lint / lint (push) Successful in 2m24s
test / unit (push) Successful in 59s
test / integration (push) Successful in 26s
test / coverage (push) Successful in 1m17s
Update Quality Badges / update-badges (push) Successful in 1m13s
manifest.py imported the extends/loader resolvers, while those resolvers
needed ManifestBottle back from manifest.py — a true bidirectional cycle
papered over with in-function imports and TYPE_CHECKING guards (not clear
dependency inversion).

Extract ManifestBottle into a new leaf module manifest_bottle.py that depends
only on the other leaf modules (manifest_util/agent/egress/git/schema).
manifest.py re-exports ManifestBottle, so `from .manifest import ManifestBottle`
callers are unaffected. With the cycle gone:

- manifest_extends and manifest_loader import ManifestBottle from
  manifest_bottle and their other deps from the real source modules, all at
  top level (TYPE_CHECKING block removed).
- manifest.py imports the extends/loader/schema/yaml_subset/log helpers at
  module top; all per-function lazy imports in the cluster are removed.

No behavior change; full unit suite green, pyright clean.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NkwFXLFff9PYPy4wgVBJp9
2026-06-26 23:42:03 -04:00
didericis-claude a256e5762a Merge pull request 'DLP injection-check perf, bounded variant cache, dedup supervise schema' (#312) from dlp-supervise-quality-fixes into main
lint / lint (push) Successful in 2m22s
test / unit (push) Successful in 50s
test / integration (push) Successful in 18s
test / coverage (push) Successful in 1m2s
Update Quality Badges / update-badges (push) Successful in 1m9s
2026-06-26 23:30:16 -04:00
5 changed files with 394 additions and 156 deletions
+12 -122
View File
@@ -62,15 +62,25 @@ from dataclasses import dataclass, field, replace
from pathlib import Path
from typing import Mapping
from .log import warn
from .manifest_util import ManifestError, as_json_object
from .manifest_agent import ManifestAgent, ManifestAgentProvider
from .manifest_bottle import ManifestBottle
from .manifest_egress import (
EGRESS_AUTH_SCHEMES,
ManifestEgressConfig,
ManifestEgressRoute,
)
from .manifest_git import ManifestGitEntry, ManifestGitUser, ManifestKeyConfig, parse_git_gate_config
from .manifest_schema import BOTTLE_KEYS
from .manifest_extends import merge_bottles_runtime, resolve_bottles
from .manifest_git import ManifestGitEntry, ManifestGitUser, ManifestKeyConfig
from .manifest_loader import (
check_stale_json,
load_bottle_chain_from_dir,
scan_agent_names,
scan_bottle_names,
)
from .manifest_schema import validate_agent_frontmatter_keys
from .yaml_subset import YamlSubsetError, parse_frontmatter
# Re-export everything that callers currently import from this module.
__all__ = [
@@ -89,10 +99,6 @@ __all__ = [
]
def _empty_str_dict() -> dict[str, str]:
return {}
def _section_dict(value: object, label: str) -> dict[str, object]:
"""Like as_json_object but treats absent/null as an empty section."""
if value is None:
@@ -100,107 +106,6 @@ def _section_dict(value: object, label: str) -> dict[str, object]:
return as_json_object(value, label)
@dataclass(frozen=True)
class ManifestBottle:
env: Mapping[str, str] = field(default_factory=_empty_str_dict)
agent_provider: ManifestAgentProvider = field(default_factory=ManifestAgentProvider)
git: tuple[ManifestGitEntry, ...] = ()
# Per-bottle git identity (issue #86). Empty default — bottles
# that don't set `git-gate.user:` in the manifest skip the
# `git config --global` step entirely. A bottle can declare a user
# identity without any git-gate.repos upstreams, and vice versa.
git_user: ManifestGitUser = field(default_factory=ManifestGitUser)
egress: ManifestEgressConfig = field(default_factory=ManifestEgressConfig)
# Per-bottle stuck-recovery sidecar (PRD 0013). When true (the
# default, issue #249), the launch step brings up a supervise
# sidecar that exposes egress MCP tools to the agent. Set
# `supervise: false` to skip the sidecar.
supervise: bool = True
@classmethod
def from_dict(cls, name: str, raw: object) -> "ManifestBottle":
d = as_json_object(raw, f"bottle '{name}'")
if "runtime" in d:
raise ManifestError(
f"bottle '{name}' has a 'runtime' field, which is no longer "
f"supported. gVisor (runsc) is now auto-detected by the "
f"backend; remove the 'runtime' field from the bottle "
f"definition."
)
if "ssh" in d:
raise ManifestError(
f"bottle '{name}' has an 'ssh' field, which has been removed "
f"(PRD 0009). Declare upstreams under 'git-gate.repos' with "
f"url + identity + host_key; the git-gate sidecar (PRD 0008) "
f"holds the credential and gitleaks-scans pushes."
)
if "git" in d:
raise ManifestError(
f"bottle '{name}' uses 'git' which has been replaced by "
f"'git-gate' (PRD 0047). Move git.user → git-gate.user "
f"and git.remotes → git-gate.repos (fields: url, identity, host_key)."
)
if "git_user" in d:
raise ManifestError(
f"bottle '{name}' has a 'git_user' field, which has been "
f"removed. Move it under 'git-gate.user'."
)
unknown = set(d.keys()) - BOTTLE_KEYS
if unknown:
allowed = ", ".join(sorted(BOTTLE_KEYS))
raise ManifestError(
f"bottle '{name}' has unknown key(s) {sorted(unknown)}; "
f"allowed keys are {allowed}."
)
env: dict[str, str] = {}
env_raw = d.get("env")
if env_raw is not None:
env_dict = as_json_object(env_raw, f"bottle '{name}' env")
for var, value in env_dict.items():
if not isinstance(value, str):
raise ManifestError(
f"env entry {var} in bottle '{name}' must be a JSON string "
f"(was {type(value).__name__}). Use \"?<message>\" for prompt-at-runtime."
)
env[var] = value
git: tuple[ManifestGitEntry, ...] = ()
git_user = ManifestGitUser()
git_raw = d.get("git-gate")
if git_raw is not None:
git, git_user = parse_git_gate_config(name, git_raw)
agent_provider = (
ManifestAgentProvider.from_dict(name, d["agent_provider"])
if "agent_provider" in d
else ManifestAgentProvider()
)
egress = (
ManifestEgressConfig.from_dict(name, d["egress"])
if "egress" in d
else ManifestEgressConfig()
)
supervise_raw = d.get("supervise", True)
if not isinstance(supervise_raw, bool):
raise ManifestError(
f"bottle '{name}' supervise must be a boolean "
f"(was {type(supervise_raw).__name__})"
)
return cls(
env=env, agent_provider=agent_provider, git=git,
git_user=git_user, egress=egress, supervise=supervise_raw,
)
def _merge_git_user(
agent_user: ManifestGitUser, base_user: ManifestGitUser
) -> ManifestGitUser:
@@ -237,8 +142,6 @@ def _resolve_effective_bottle_eager(
When bottle_names is non-empty they are merged in order. When empty, falls
back to agent.bottle. Raises ManifestError when neither is set."""
from .manifest_extends import merge_bottles_runtime
if bottle_names:
resolved: list[ManifestBottle] = []
for bn in bottle_names:
@@ -270,9 +173,6 @@ def _resolve_effective_bottle_lazy(
When bottle_names is non-empty they are resolved from disk and merged in
order. When empty, falls back to agent_bottle. Raises ManifestError when
neither is set."""
from .manifest_extends import merge_bottles_runtime
from .manifest_loader import load_bottle_chain_from_dir
if bottle_names:
resolved = [load_bottle_chain_from_dir(bn, bottles_dir) for bn in bottle_names]
return merge_bottles_runtime(resolved)
@@ -358,8 +258,6 @@ class ManifestIndex:
home_md = home_dir / ".bot-bottle"
cwd_md = cwd_dir / ".bot-bottle"
from .manifest_loader import check_stale_json
check_stale_json(home_dir, home_md, "$HOME")
if cwd_dir.resolve() != home_dir.resolve():
check_stale_json(cwd_dir, cwd_md, "$CWD")
@@ -399,7 +297,6 @@ class ManifestIndex:
files = sorted(stale_bottles.glob("*.md"))
if files:
names = ", ".join(p.name for p in files)
from .log import warn
warn(
f"ignoring bottle file(s) under "
f"{stale_bottles}: {names}. Bottles can only "
@@ -421,7 +318,6 @@ class ManifestIndex:
raw_bottles: dict[str, dict[str, object]] = {}
for n, b in raw_bottles_obj.items():
raw_bottles[n] = as_json_object(b, f"bottle '{n}'")
from .manifest_extends import resolve_bottles
bottles = resolve_bottles(raw_bottles)
@@ -439,7 +335,6 @@ class ManifestIndex:
filenames without reading their content. In eager mode (from
from_json_obj) it returns the pre-parsed bottles' names."""
if self.home_md is not None:
from .manifest_loader import scan_bottle_names
return scan_bottle_names(self.home_md / "bottles")
return sorted(self.bottles.keys())
@@ -451,7 +346,6 @@ class ManifestIndex:
filenames without reading their content. In eager mode (from
from_json_obj) it returns the pre-parsed agents' names."""
if self.home_md is not None:
from .manifest_loader import scan_agent_names
home_names = set(scan_agent_names(self.home_md / "agents").keys())
cwd_names: set[str] = set()
if self.cwd_md is not None:
@@ -509,10 +403,6 @@ class ManifestIndex:
"""Lazy path (resolve/from_md_dirs): read and parse the agent file and
its bottle chain from disk for the first time here."""
assert self.home_md is not None # guaranteed by load_for_agent dispatch
from .manifest_loader import scan_agent_names
from .manifest_schema import validate_agent_frontmatter_keys
from .yaml_subset import YamlSubsetError, parse_frontmatter
# Locate the agent file; cwd wins over home on name collision.
home_agents = scan_agent_names(self.home_md / "agents")
cwd_agents: dict[str, Path] = {}
+129
View File
@@ -0,0 +1,129 @@
"""The `ManifestBottle` value type.
Split out of `manifest.py` so the `extends:`/loader resolvers can import it
without a circular dependency: `manifest.py` imports those resolvers, while
they only need this value type. Everything here depends on leaf modules
(`manifest_util`, `manifest_agent`, `manifest_egress`, `manifest_git`,
`manifest_schema`), so this module sits at the bottom of the manifest layer.
`manifest.py` re-exports `ManifestBottle`, so existing
`from .manifest import ManifestBottle` callers are unaffected.
"""
from __future__ import annotations
from dataclasses import dataclass, field
from typing import Mapping
from .manifest_util import ManifestError, as_json_object
from .manifest_agent import ManifestAgentProvider
from .manifest_egress import ManifestEgressConfig
from .manifest_git import ManifestGitEntry, ManifestGitUser, parse_git_gate_config
from .manifest_schema import BOTTLE_KEYS
__all__ = ["ManifestBottle"]
def _empty_str_dict() -> dict[str, str]:
return {}
@dataclass(frozen=True)
class ManifestBottle:
env: Mapping[str, str] = field(default_factory=_empty_str_dict)
agent_provider: ManifestAgentProvider = field(default_factory=ManifestAgentProvider)
git: tuple[ManifestGitEntry, ...] = ()
# Per-bottle git identity (issue #86). Empty default — bottles
# that don't set `git-gate.user:` in the manifest skip the
# `git config --global` step entirely. A bottle can declare a user
# identity without any git-gate.repos upstreams, and vice versa.
git_user: ManifestGitUser = field(default_factory=ManifestGitUser)
egress: ManifestEgressConfig = field(default_factory=ManifestEgressConfig)
# Per-bottle stuck-recovery sidecar (PRD 0013). When true (the
# default, issue #249), the launch step brings up a supervise
# sidecar that exposes egress MCP tools to the agent. Set
# `supervise: false` to skip the sidecar.
supervise: bool = True
@classmethod
def from_dict(cls, name: str, raw: object) -> "ManifestBottle":
d = as_json_object(raw, f"bottle '{name}'")
if "runtime" in d:
raise ManifestError(
f"bottle '{name}' has a 'runtime' field, which is no longer "
f"supported. gVisor (runsc) is now auto-detected by the "
f"backend; remove the 'runtime' field from the bottle "
f"definition."
)
if "ssh" in d:
raise ManifestError(
f"bottle '{name}' has an 'ssh' field, which has been removed "
f"(PRD 0009). Declare upstreams under 'git-gate.repos' with "
f"url + identity + host_key; the git-gate sidecar (PRD 0008) "
f"holds the credential and gitleaks-scans pushes."
)
if "git" in d:
raise ManifestError(
f"bottle '{name}' uses 'git' which has been replaced by "
f"'git-gate' (PRD 0047). Move git.user → git-gate.user "
f"and git.remotes → git-gate.repos (fields: url, identity, host_key)."
)
if "git_user" in d:
raise ManifestError(
f"bottle '{name}' has a 'git_user' field, which has been "
f"removed. Move it under 'git-gate.user'."
)
unknown = set(d.keys()) - BOTTLE_KEYS
if unknown:
allowed = ", ".join(sorted(BOTTLE_KEYS))
raise ManifestError(
f"bottle '{name}' has unknown key(s) {sorted(unknown)}; "
f"allowed keys are {allowed}."
)
env: dict[str, str] = {}
env_raw = d.get("env")
if env_raw is not None:
env_dict = as_json_object(env_raw, f"bottle '{name}' env")
for var, value in env_dict.items():
if not isinstance(value, str):
raise ManifestError(
f"env entry {var} in bottle '{name}' must be a JSON string "
f"(was {type(value).__name__}). Use \"?<message>\" for prompt-at-runtime."
)
env[var] = value
git: tuple[ManifestGitEntry, ...] = ()
git_user = ManifestGitUser()
git_raw = d.get("git-gate")
if git_raw is not None:
git, git_user = parse_git_gate_config(name, git_raw)
agent_provider = (
ManifestAgentProvider.from_dict(name, d["agent_provider"])
if "agent_provider" in d
else ManifestAgentProvider()
)
egress = (
ManifestEgressConfig.from_dict(name, d["egress"])
if "egress" in d
else ManifestEgressConfig()
)
supervise_raw = d.get("supervise", True)
if not isinstance(supervise_raw, bool):
raise ManifestError(
f"bottle '{name}' supervise must be a boolean "
f"(was {type(supervise_raw).__name__})"
)
return cls(
env=env, agent_provider=agent_provider, git=git,
git_user=git_user, egress=egress, supervise=supervise_raw,
)
+4 -28
View File
@@ -2,11 +2,10 @@
from __future__ import annotations
from typing import TYPE_CHECKING
if TYPE_CHECKING:
from .manifest import ManifestBottle
from .manifest_egress import ManifestEgressConfig
from .manifest_bottle import ManifestBottle
from .manifest_egress import ManifestEgressConfig, validate_egress_routes
from .manifest_git import ManifestGitUser, parse_git_gate_config
from .manifest_util import ManifestError, as_json_object
def merge_bottles_runtime(bottles: "list[ManifestBottle]") -> "ManifestBottle":
@@ -27,9 +26,6 @@ def merge_bottles_runtime(bottles: "list[ManifestBottle]") -> "ManifestBottle":
def _merge_two_bottles_runtime(base: "ManifestBottle", override: "ManifestBottle") -> "ManifestBottle":
from .manifest import ManifestBottle, ManifestGitUser
from .manifest_egress import ManifestEgressConfig
merged_env = {**base.env, **override.env}
merged_git_user = ManifestGitUser(
@@ -81,8 +77,6 @@ def _resolve_one_bottle(
repos_cache: dict[str, dict[str, object]],
seen: tuple[str, ...],
) -> ManifestBottle:
from .manifest import ManifestBottle, ManifestError
if name in cache:
return cache[name]
if name in seen:
@@ -174,11 +168,6 @@ def _fold_two_bottles(
later_repos_raw: dict[str, object],
) -> tuple[ManifestBottle, dict[str, object]]:
"""Combine two resolved parent bottles; later wins over earlier."""
from .manifest import ManifestBottle, ManifestGitUser
from .manifest_egress import ManifestEgressConfig
from .manifest_git import parse_git_gate_config
from .manifest_util import as_json_object
merged_env = {**earlier.env, **later.env}
merged_git_user = ManifestGitUser(
@@ -227,10 +216,6 @@ def _merge_bottles(
name: str,
) -> ManifestBottle:
"""Apply PRD 0025 merge rules."""
from .manifest import ManifestBottle, ManifestGitUser
from .manifest_egress import validate_egress_routes
from .manifest_util import as_json_object
# git-gate.repos: when the child declares repos, inject the already
# name-merged repo set (computed by _resolve_repos_raw) so the child
# parses with the full inherited+overridden list (issue #237).
@@ -303,8 +288,6 @@ def _resolve_repos_raw(
inherits the parent's set verbatim; an explicit empty dict clears it.
Otherwise parent and child unite by name, with same-name entries
field-merged (parent fields are defaults, child fields win)."""
from .manifest_util import as_json_object
if not _child_declares_git_gate_repos(child_raw):
return parent_repos
child_repos = _declared_repos_raw(child_raw)
@@ -324,8 +307,6 @@ def _resolve_repos_raw(
def _declared_repos_raw(child_raw: dict[str, object]) -> dict[str, object]:
"""Return the child's explicitly declared git-gate.repos as raw dicts,
or an empty dict when none are declared."""
from .manifest_util import as_json_object
if not _child_declares_git_gate_repos(child_raw):
return {}
git_raw = as_json_object(child_raw.get("git-gate", {}), "child git-gate")
@@ -333,8 +314,6 @@ def _declared_repos_raw(child_raw: dict[str, object]) -> dict[str, object]:
def _child_declares_git_gate_repos(child_raw: dict[str, object]) -> bool:
from .manifest_util import as_json_object
git_raw = child_raw.get("git-gate")
if git_raw is None:
return False
@@ -347,9 +326,6 @@ def _merge_egress(
child: ManifestEgressConfig,
child_raw: dict[str, object],
) -> ManifestEgressConfig:
from .manifest_egress import ManifestEgressConfig
from .manifest_util import as_json_object
child_egress_raw = as_json_object(child_raw.get("egress"), "child egress")
routes = parent.routes + child.routes
log = child.Log if "log" in child_egress_raw else parent.Log
+2 -6
View File
@@ -3,9 +3,10 @@
from __future__ import annotations
from pathlib import Path
from typing import TYPE_CHECKING
from .log import warn
from .manifest_bottle import ManifestBottle
from .manifest_extends import resolve_bottles
from .manifest_schema import (
entity_name_from_path,
validate_bottle_frontmatter_keys,
@@ -13,9 +14,6 @@ from .manifest_schema import (
from .manifest_util import ManifestError
from .yaml_subset import YamlSubsetError, parse_frontmatter
if TYPE_CHECKING:
from .manifest import ManifestBottle
def check_stale_json(dir_path: Path, md_dir: Path, label: str) -> None:
"""Die if `<dir_path>/bot-bottle.json` exists but `md_dir` does
@@ -78,8 +76,6 @@ def load_bottle_chain_from_dir(
Only the files in the extends chain are read — unrelated bottle files
are never touched. Raises ManifestError on parse or validation failure."""
from .manifest_extends import resolve_bottles
raws: dict[str, dict[str, object]] = {}
to_load = [bottle_name]
while to_load:
+247
View File
@@ -0,0 +1,247 @@
# PRD prd-new: Egress control plane — metering, budgets, and forced cutoff
- **Status:** Draft
- **Author:** didericis
- **Created:** 2026-06-25
- **Issue:** #251
## Summary
Add an **out-of-band egress enforcement & observability plane**: meter every
agent's token usage at the egress proxy, decrement budgets without the agent's
cooperation, and forcibly cut a bottle's egress when a budget is exhausted —
either automatically or on command from a host-level dashboard. The trigger
(usage threshold) and the action (route-drop / freeze / kill) both live in the
egress plane and run with no agent in the loop. This is distinct from the
supervise sidecar (PRD 0013), which is agent-initiated and therefore cannot
enforce a cost cutoff on a runaway agent. State (usage ledger, budgets, audit)
moves into a host-level SQLite database behind a thin repository API, the first
SQL store in an otherwise flat-file repo.
## Problem
bot-bottle can't currently do two things the cost-overrun case demands:
1. **Forced egress shutdown on limit.** When an agent crosses a token
threshold, kill its egress automatically — no human in the loop.
2. **Remote (host-level) management.** Drive agents from a single surface:
see usage, cut egress, stop bottles, to prevent cost overruns.
The existing supervise sidecar (PRD 0013) is **entirely agent-initiated**: every
action begins with the agent voluntarily calling an MCP tool and an operator
approving it. A runaway or expensive agent — exactly the cost-overrun case —
will never call `egress-block` on itself. Supervision is therefore a
**collaborative recovery** mechanism, not an **enforcement** mechanism; making
it mandatory (#249) would not deliver forced cost-cutoff.
The requirement forces a distinction the current design blurs:
- **Plane A — enforcement / observability (this PRD).** System → infrastructure.
Meter usage, cut egress on threshold or command, account for cost.
Out-of-band; independent of the agent. **Unconditional** — an enforcement
plane you can opt out of isn't enforcement.
- **Plane B — agent-facing recovery (the existing supervise sidecar).**
Agent → operator, approval-gated. Useful interactively; meaningless for a
headless agent with no operator watching its queue. Remains optional.
This PRD builds Plane A. It reframes the "always-on control" invariant of #249
as "the egress control plane is always present" — a more defensible property
than "every agent runs the agent-facing supervisor." Unsupervised
(headless/CI/ephemeral) agents stay first-class: still subject to the mandatory
meter + kill switch, they simply lack the agent-facing proposal tools they
couldn't use anyway.
## Goals / Success Criteria
- The egress proxy meters every request to a metered API host (e.g.
`api.anthropic.com`) and records authoritative token usage per bottle and per
agent provider, with no agent cooperation.
- A budget can be set at four scopes with deterministic precedence
(**agent → bottle → parent bottle → global host budget**); the
most-specific applicable budget governs.
- When usage crosses a budget, the bottle's configured **cutoff policy**
(`cutoff` | `freeze` | `kill`) fires automatically, executed host-side on the
egress plane — never via the supervise queue.
- An operator can, from a single **host-level TUI dashboard**, see live per-bottle
usage against budget and command a cutoff/stop on demand.
- Host budgets, default cutoff policy, and per-provider limits are declared in a
new host-level `~/.bot-bottle/settings.yml`, parseable by `yaml_subset.py`.
- All usage, budget state, and enforcement actions persist in a host-level
SQLite DB behind a thin repository API, so the store can later be swapped for
a cross-host cloud service.
## Non-goals
- **Remote control / cross-host control plane.** Web + mobile remote control,
cross-host budgets, and the authn/transport they require are explicitly
deferred. v1 is a **host-only TUI** with no remote surface.
- **Dollar-denominated budgets.** Budgets are token counts keyed by agent
provider, not currency. Price tables are out of scope.
- **Migrating existing flat-file state into SQLite.** Resume `metadata.json`,
transcripts, Dockerfile overrides, the supervise queue, and audit logs stay on
the filesystem. Only the *new* metering/budget/enforcement ledger is SQL.
- **Making the supervise sidecar (Plane B) mandatory.** Out of scope here; this
PRD is the answer to "what should be unconditional" (Plane A), leaving #249's
Plane-B question open.
- **Per-request hard pre-send blocking as the primary mechanism.** The gate is
budget-crossing detected at/after metering; a pre-flight estimator (below) is a
refinement, not the core enforcement path.
## Design
### Two measurements: gate vs. account
There are two distinct needs, and they want different signals:
- **Account (authoritative).** Decrement the real budget from the API
**response**, which already carries authoritative usage (Anthropic
`input_tokens` / `output_tokens`, OpenAI `usage`). The egress addon already
has a `response(flow)` hook (`bot_bottle/egress_addon.py:460`), so the real
number is available with no extra network call. **Caveat:** agent traffic is
mostly streaming SSE, so the response path must tail the stream for the final
usage event rather than parse a single JSON body — scoped explicitly as work.
- **Gate (estimate).** To block *before* sending, only the request is available,
so an estimator / provider `count_tokens` endpoint is the only option.
Calling `count_tokens` for accounting would be both less accurate *and* an extra
metered egress call per request, so accounting uses response `usage` and the
estimator is reserved for the optional pre-flight gate.
### `count_tokens` on agent providers
Add an abstract `count_tokens(request) -> int` to the `AgentProvider`
abstraction (`bot_bottle/agent_provider.py`):
- **Default** is a good-enough stdlib estimator. Prefer stdlib only; a small
pip dependency *for the sidecar* is acceptable for the fallback if stdlib
proves too inaccurate (this does not relax the package's stdlib-first stance —
it would be a sidecar-only dep, like the bundle already carries).
- **Built-in `claude`** uses Anthropic's token-counting endpoint;
**built-in `codex`** uses OpenAI's. These are exact for the gate but cost a
metered call, so they are gate-only; accounting still comes from the response.
### Budgets and precedence
Budgets are token counts keyed by **agent provider name** (the same names
bottles already use). Four scopes, most-specific wins:
```
agent → bottle → parent bottle → global (host)
```
The global host budget is the highest-priority feature to ship (the cross-host
control plane will eventually consume it); per-agent and per-bottle budgets
override it for finer control. A budget can also be supplied **at bottle
launch** (`--budget` or equivalent), overriding the settings.yml defaults for
that run. Enforcement evaluates the effective budget as the
nearest-defined scope at decrement time.
### `~/.bot-bottle/settings.yml`
New **host-level** settings file (the `~/.bot-bottle/` root, *not* the per-repo
`.bot-bottle/` — host budgets must not be committed per-repo). Parsed by
`yaml_subset.py`, so it must stay within that bounded subset (flat mappings,
scalars; no anchors, no multi-line block scalars). Shape:
```yaml
budget:
claude: 5000000 # token budget keyed by agent provider
codex: 2000000
shutdown: cutoff # default cutoff policy: cutoff | freeze | kill
```
### Forced cutoff and cutoff policy
On budget exhaustion (or an operator command), the configured per-bottle cutoff
policy fires. The three policies map onto primitives that already exist:
- **`cutoff`** (default) — drop the bottle's `routes.yaml` to empty and reload
(or isolate the bottle from the egress network); the agent/bottle keeps
running but can no longer reach metered hosts. This is the route-drop already
available on the egress plane (`bot_bottle/backend/egress_apply.py`).
- **`freeze`** — commit/snapshot state, then kill the agent/bottle; resumable
later via `bot_bottle/backend/freeze.py`.
- **`kill`** — tear the bottle down without saving state (backend teardown).
The trigger lives in the metering path and the action in the egress/backend
plane; **neither touches the supervise proposal queue** (design constraint from
#251).
### Host-level SQLite store
**Decision: introduce SQLite now, narrowly.**
- **The dependency objection doesn't apply.** `sqlite3` is in the Python stdlib,
so it does not break the AGENTS.md stdlib-first / no-runtime-pip stance — same
category as the hand-rolled `yaml_subset.py`, except the stdlib already ships
the whole engine.
- **It fits the problem.** A *global* token budget decremented concurrently by N
egress sidecars (today `~/.bot-bottle/` already has `state/`, `audit/`,
`queue/` written by parallel bottles) is a read-modify-write race. Over JSON
that means hand-rolled file locking; SQLite gives atomic transactions + WAL for
free. The per-agent/per-bottle precedence rollup plus "sum across all bottles"
is a `GROUP BY`, not an N-directory rescan.
- **It rehearses the cloud swap.** "Wrap operations in an API so we can swap to a
cloud service" maps directly onto a thin repository/DAO over SQLite → Postgres
later. A JSON-file store is a worse rehearsal than SQL.
**Costs (real but bounded):** a new paradigm in a flat-file repo needs a
`schema_version` table + idempotent startup migrations; SQLite serializes
writers, so WAL mode + `busy_timeout` are required (a non-issue at a handful of
bottles); test fixtures need temp DBs.
**Scope of the store:** one DB at `~/.bot-bottle/bot-bottle.db` behind a thin
repository API. Only the **new** metering/budget/enforcement-audit ledger lives
there. Existing per-bottle blobs (resume `metadata.json`, transcripts,
Dockerfile overrides, supervise queue) stay on the filesystem — migrating them
now is churn for no benefit and they lack the concurrency/aggregation problem.
### Host-level controller + dashboard
A single **host-level controller** owns the meter, budget evaluation, and the
cutoff actions across all bottles (cf. `bot_bottle/cli/supervise.py`'s
cross-bottle view), rather than a per-bottle daemon. v1 ships one host-level
**TUI dashboard** that reads live usage-vs-budget from the SQLite store and
offers on-demand cutoff/stop. The existing supervisor UI should eventually fold
into this same dashboard; this PRD lays the host-level surface it will move to.
## Implementation chunks
Ordered, individually mergeable:
1. **SQLite repository foundation.** `~/.bot-bottle/bot-bottle.db`, schema +
`schema_version` migrations, WAL + `busy_timeout`, thin repository API,
temp-DB test fixtures. No behavior wired yet.
2. **Metering at the egress proxy.** Parse authoritative response `usage`
(including SSE final-usage tailing) in the egress addon `response` hook;
write per-bottle / per-provider usage rows to the ledger.
3. **`settings.yml` + budget model.** Host-level `~/.bot-bottle/settings.yml`
parsed by `yaml_subset.py`; budget precedence (agent → bottle → parent →
global) and the `--budget` launch flag.
4. **Forced cutoff + cutoff policy.** Wire the threshold trigger to the
`cutoff` / `freeze` / `kill` primitives on the egress/backend plane; record
enforcement actions to the audit ledger.
5. **Host-level TUI dashboard.** Live usage-vs-budget view + on-demand
cutoff/stop, reading the store.
6. **`count_tokens` pre-flight gate (optional refinement).** Abstract method +
stdlib estimator default; Anthropic/OpenAI endpoints for built-in
claude/codex; optional pre-send block.
## Open questions
- **SSE usage tailing robustness.** Buffering streamed responses to extract the
final usage event without breaking the agent's own stream consumption — how
much of the body must the addon hold, and what's the failure mode if the
stream is interrupted mid-flight?
- **Crossing mid-request.** A single response can push usage past budget only
*after* it's already been delivered. Is post-hoc cutoff (next request blocked)
sufficient, or is a pre-flight estimator gate (chunk 6) required for v1?
- **Provider name ↔ metered host mapping.** How does the proxy attribute a
flow to an agent-provider budget key — by destination host, by bottle
identity, or both?
- **Parent-bottle budget semantics.** For `bottle extends` (PRD 0025 / 0065)
chains, does "parent bottle" mean the manifest parent, the launching bottle,
or the full ancestry summed?
- **Dashboard ↔ controller transport (even host-only).** In-process, a local
socket, or polling the SQLite store directly? Picks the seam the future remote
control plane will extend.