feat: pluggable bootstrap framework with ordered initialisers (#847)

A generic, long-running bootstrap processor that converges a
deployment to its configured initial state and then idles.
Replaces the previous one-shot `tg-init-trustgraph` container model
and provides an extension point for enterprise / third-party
initialisers.

See docs/tech-specs/bootstrap.md for the full design.

Bootstrapper
------------
A single AsyncProcessor (trustgraph.bootstrap.bootstrapper.Processor)
that:

  * Reads a list of initialiser specifications (class, name, flag,
    params) from either a direct `initialisers` parameter
    (processor-group embedding) or a YAML/JSON file (`-c`, CLI).
  * On each wake, runs a cheap service-gate (config-svc +
    flow-svc round-trips), then iterates the initialiser list,
    running each whose configured flag differs from the one stored
    in __system__/init-state/<name>.
  * Stores per-initialiser completion state in the reserved
    __system__ workspace.
  * Adapts cadence: ~5s on gate failure, ~15s while converging,
    ~300s in steady state.
  * Isolates failures — one initialiser's exception does not block
    others in the same cycle; the failed one retries next wake.

Initialiser contract
--------------------
  * Subclass trustgraph.bootstrap.base.Initialiser.
  * Implement async run(ctx, old_flag, new_flag).
  * Opt out of the service gate with class attr
    wait_for_services=False (only used by PulsarTopology, since
    config-svc cannot come up until Pulsar namespaces exist).
  * ctx carries short-lived config and flow-svc clients plus a
    scoped logger.

Core initialisers (trustgraph.bootstrap.initialisers.*)
-------------------------------------------------------
  * PulsarTopology   — creates Pulsar tenant + namespaces
                       (pre-gate, blocking HTTP offloaded to
                        executor).
  * TemplateSeed     — seeds __template__ from an external JSON
                       file; re-run is upsert-missing by default,
                       overwrite-all opt-in.
  * WorkspaceInit    — populates a named workspace from either
                       the full contents of __template__ or a
                       seed file; raises cleanly if the template
                       isn't seeded yet so the bootstrapper retries
                       on the next cycle.
  * DefaultFlowStart — starts a specific flow in a workspace;
                       no-ops if the flow is already running.

Enterprise or third-party initialisers plug in via fully-qualified
dotted class paths in the bootstrapper's configuration — no core
code change required.

Config service
--------------
  * push(): filter out reserved workspaces (ids starting with "_")
    from the change notifications.  Stored config is preserved; only
    the broadcast is suppressed, so bootstrap / template state lives
    in config-svc without live processors ever reacting to it.

Config client
-------------
  * ConfigClient.get_all(workspace): wraps the existing `config`
    operation to return {type: {key: value}} for a workspace.
    WorkspaceInit uses it to copy __template__ without needing a
    hardcoded types list.

pyproject.toml
--------------
  * Adds a `bootstrap` console script pointing at the new Processor.

* Remove tg-init-trustgraph, superceded by bootstrap processor
This commit is contained in:
cybermaggedon 2026-04-22 18:03:46 +01:00 committed by GitHub
parent 31027e30ae
commit ae9936c9cc
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
17 changed files with 1312 additions and 273 deletions

View file

@ -0,0 +1,20 @@
"""
Core bootstrap initialisers.
These cover the base TrustGraph deployment case. Enterprise or
third-party initialisers live in their own packages and are
referenced in the bootstrapper's config by fully-qualified dotted
path.
"""
from . pulsar_topology import PulsarTopology
from . template_seed import TemplateSeed
from . workspace_init import WorkspaceInit
from . default_flow_start import DefaultFlowStart
__all__ = [
"PulsarTopology",
"TemplateSeed",
"WorkspaceInit",
"DefaultFlowStart",
]

View file

@ -0,0 +1,101 @@
"""
DefaultFlowStart initialiser starts a named flow in a workspace
using a specified blueprint.
Separated from WorkspaceInit so deployments that want a workspace
without an auto-started flow can simply omit this initialiser.
Parameters
----------
workspace : str (default "default")
Workspace in which to start the flow.
flow_id : str (default "default")
Identifier for the started flow.
blueprint : str (required)
Blueprint name (must already exist in the workspace's config,
typically via TemplateSeed -> WorkspaceInit).
description : str (default "Default")
Human-readable description passed to flow-svc.
parameters : dict (optional)
Optional parameter overrides passed to start-flow.
"""
from trustgraph.schema import FlowRequest
from .. base import Initialiser
class DefaultFlowStart(Initialiser):
def __init__(
self,
workspace="default",
flow_id="default",
blueprint=None,
description="Default",
parameters=None,
**kwargs,
):
super().__init__(**kwargs)
if not blueprint:
raise ValueError(
"DefaultFlowStart requires 'blueprint'"
)
self.workspace = workspace
self.flow_id = flow_id
self.blueprint = blueprint
self.description = description
self.parameters = dict(parameters) if parameters else {}
async def run(self, ctx, old_flag, new_flag):
# Check whether the flow already exists. Belt-and-braces
# beyond the flag gate: if an operator stops and restarts the
# bootstrapper after the flow is already running, we don't
# want to blindly try to start it again.
list_resp = await ctx.flow.request(
FlowRequest(
operation="list-flows",
workspace=self.workspace,
),
timeout=10,
)
if list_resp.error:
raise RuntimeError(
f"list-flows failed: "
f"{list_resp.error.type}: {list_resp.error.message}"
)
if self.flow_id in (list_resp.flow_ids or []):
ctx.logger.info(
f"Flow {self.flow_id!r} already running in workspace "
f"{self.workspace!r}; nothing to do"
)
return
ctx.logger.info(
f"Starting flow {self.flow_id!r} "
f"(blueprint={self.blueprint!r}) "
f"in workspace {self.workspace!r}"
)
resp = await ctx.flow.request(
FlowRequest(
operation="start-flow",
workspace=self.workspace,
flow_id=self.flow_id,
blueprint_name=self.blueprint,
description=self.description,
parameters=self.parameters,
),
timeout=30,
)
if resp.error:
raise RuntimeError(
f"start-flow failed: "
f"{resp.error.type}: {resp.error.message}"
)
ctx.logger.info(
f"Flow {self.flow_id!r} started"
)

View file

@ -0,0 +1,131 @@
"""
PulsarTopology initialiser creates Pulsar tenant and namespaces
with their retention policies.
Runs pre-gate (``wait_for_services = False``) because config-svc and
flow-svc can't connect to Pulsar until these namespaces exist.
Admin-API calls are idempotent so re-runs on flag change are safe.
"""
import asyncio
import requests
from .. base import Initialiser
# Namespace configs. flow/request take broker defaults. response
# and notify get aggressive retention — those classes carry short-lived
# request/response and notification traffic only.
NAMESPACE_CONFIG = {
"flow": {},
"request": {},
"response": {
"retention_policies": {
"retentionSizeInMB": -1,
"retentionTimeInMinutes": 3,
"subscriptionExpirationTimeMinutes": 30,
},
},
"notify": {
"retention_policies": {
"retentionSizeInMB": -1,
"retentionTimeInMinutes": 3,
"subscriptionExpirationTimeMinutes": 5,
},
},
}
REQUEST_TIMEOUT = 10
class PulsarTopology(Initialiser):
wait_for_services = False
def __init__(
self,
admin_url="http://pulsar:8080",
tenant="tg",
**kwargs,
):
super().__init__(**kwargs)
self.admin_url = admin_url.rstrip("/")
self.tenant = tenant
async def run(self, ctx, old_flag, new_flag):
# requests is blocking; offload to executor so the loop stays
# responsive.
loop = asyncio.get_event_loop()
await loop.run_in_executor(None, self._reconcile_sync, ctx.logger)
# ------------------------------------------------------------------
# Sync admin-API calls.
# ------------------------------------------------------------------
def _get_clusters(self):
resp = requests.get(
f"{self.admin_url}/admin/v2/clusters",
timeout=REQUEST_TIMEOUT,
)
resp.raise_for_status()
return resp.json()
def _tenant_exists(self):
resp = requests.get(
f"{self.admin_url}/admin/v2/tenants/{self.tenant}",
timeout=REQUEST_TIMEOUT,
)
return resp.status_code == 200
def _create_tenant(self, clusters):
resp = requests.put(
f"{self.admin_url}/admin/v2/tenants/{self.tenant}",
json={"adminRoles": [], "allowedClusters": clusters},
timeout=REQUEST_TIMEOUT,
)
if resp.status_code != 204:
raise RuntimeError(
f"Tenant {self.tenant!r} create failed: "
f"{resp.status_code} {resp.text}"
)
def _namespace_exists(self, namespace):
resp = requests.get(
f"{self.admin_url}/admin/v2/namespaces/"
f"{self.tenant}/{namespace}",
timeout=REQUEST_TIMEOUT,
)
return resp.status_code == 200
def _create_namespace(self, namespace, config):
resp = requests.put(
f"{self.admin_url}/admin/v2/namespaces/"
f"{self.tenant}/{namespace}",
json=config,
timeout=REQUEST_TIMEOUT,
)
if resp.status_code != 204:
raise RuntimeError(
f"Namespace {self.tenant}/{namespace} create failed: "
f"{resp.status_code} {resp.text}"
)
def _reconcile_sync(self, logger):
if not self._tenant_exists():
clusters = self._get_clusters()
logger.info(
f"Creating tenant {self.tenant!r} with clusters {clusters}"
)
self._create_tenant(clusters)
else:
logger.debug(f"Tenant {self.tenant!r} already exists")
for namespace, config in NAMESPACE_CONFIG.items():
if self._namespace_exists(namespace):
logger.debug(
f"Namespace {self.tenant}/{namespace} already exists"
)
continue
logger.info(
f"Creating namespace {self.tenant}/{namespace}"
)
self._create_namespace(namespace, config)

View file

@ -0,0 +1,93 @@
"""
TemplateSeed initialiser populates the reserved ``__template__``
workspace from an external JSON seed file.
Seed file shape:
.. code-block:: json
{
"flow-blueprint": {
"ontology": { ... },
"agent": { ... }
},
"prompt": {
...
},
...
}
Top-level keys are config types; nested keys are config entries.
Values are arbitrary JSON (they'll be ``json.dumps()``'d on write).
Parameters
----------
config_file : str
Path to the seed file on disk.
overwrite : bool (default False)
On re-run (flag change), if True overwrite all keys; if False
upsert-missing-only (preserves any operator customisation of
the template).
"""
import json
from .. base import Initialiser
TEMPLATE_WORKSPACE = "__template__"
class TemplateSeed(Initialiser):
def __init__(self, config_file, overwrite=False, **kwargs):
super().__init__(**kwargs)
if not config_file:
raise ValueError("TemplateSeed requires 'config_file'")
self.config_file = config_file
self.overwrite = overwrite
async def run(self, ctx, old_flag, new_flag):
with open(self.config_file) as f:
seed = json.load(f)
if old_flag is None:
# Clean first run — write every entry.
await self._write_all(ctx, seed)
return
# Re-run after flag change.
if self.overwrite:
await self._write_all(ctx, seed)
else:
await self._upsert_missing(ctx, seed)
async def _write_all(self, ctx, seed):
values = []
for type_name, entries in seed.items():
for key, value in entries.items():
values.append((type_name, key, json.dumps(value)))
if values:
await ctx.config.put_many(TEMPLATE_WORKSPACE, values)
ctx.logger.info(
f"Template seeded with {len(values)} entries"
)
async def _upsert_missing(self, ctx, seed):
written = 0
for type_name, entries in seed.items():
existing = set(
await ctx.config.keys(TEMPLATE_WORKSPACE, type_name)
)
values = []
for key, value in entries.items():
if key not in existing:
values.append(
(type_name, key, json.dumps(value))
)
if values:
await ctx.config.put_many(TEMPLATE_WORKSPACE, values)
written += len(values)
ctx.logger.info(
f"Template upsert-missing: {written} new entries"
)

View file

@ -0,0 +1,138 @@
"""
WorkspaceInit initialiser creates a workspace and populates it from
either the ``__template__`` workspace or a seed file on disk.
Parameters
----------
workspace : str
Target workspace to create / populate.
source : str
Either ``"template"`` (copy the full contents of the
``__template__`` workspace) or ``"seed-file"`` (read from
``seed_file``).
seed_file : str (required when source=="seed-file")
Path to a JSON seed file with the same shape TemplateSeed consumes.
overwrite : bool (default False)
On re-run (flag change), if True overwrite all keys; if False,
upsert-missing-only (preserves in-workspace customisations).
Raises (in ``run``)
-------------------
When source is ``"template"``, raises ``RuntimeError`` if the
``__template__`` workspace is empty indicating that TemplateSeed
hasn't run yet. The bootstrapper's retry loop will re-attempt on
the next cycle once the prerequisite is satisfied.
"""
import json
from .. base import Initialiser
TEMPLATE_WORKSPACE = "__template__"
class WorkspaceInit(Initialiser):
def __init__(
self,
workspace="default",
source="template",
seed_file=None,
overwrite=False,
**kwargs,
):
super().__init__(**kwargs)
if source not in ("template", "seed-file"):
raise ValueError(
f"WorkspaceInit: source must be 'template' or "
f"'seed-file', got {source!r}"
)
if source == "seed-file" and not seed_file:
raise ValueError(
"WorkspaceInit: seed_file required when source='seed-file'"
)
self.workspace = workspace
self.source = source
self.seed_file = seed_file
self.overwrite = overwrite
async def run(self, ctx, old_flag, new_flag):
if self.source == "seed-file":
tree = self._load_seed_file()
else:
tree = await self._load_from_template(ctx)
if old_flag is None or self.overwrite:
await self._write_all(ctx, tree)
else:
await self._upsert_missing(ctx, tree)
def _load_seed_file(self):
with open(self.seed_file) as f:
return json.load(f)
async def _load_from_template(self, ctx):
"""Build a seed tree from the entire ``__template__`` workspace.
Raises if the workspace is empty, so the bootstrapper knows
the prerequisite isn't met yet."""
raw_tree = await ctx.config.get_all(TEMPLATE_WORKSPACE)
tree = {}
total = 0
for type_name, entries in raw_tree.items():
parsed = {}
for key, raw in entries.items():
if raw is None:
continue
try:
parsed[key] = json.loads(raw)
except Exception:
parsed[key] = raw
total += 1
if parsed:
tree[type_name] = parsed
if total == 0:
raise RuntimeError(
"Template workspace is empty — has TemplateSeed run yet?"
)
ctx.logger.debug(
f"Loaded {total} template entries across {len(tree)} types"
)
return tree
async def _write_all(self, ctx, tree):
values = []
for type_name, entries in tree.items():
for key, value in entries.items():
values.append((type_name, key, json.dumps(value)))
if values:
await ctx.config.put_many(self.workspace, values)
ctx.logger.info(
f"Workspace {self.workspace!r} populated with "
f"{len(values)} entries"
)
async def _upsert_missing(self, ctx, tree):
written = 0
for type_name, entries in tree.items():
existing = set(
await ctx.config.keys(self.workspace, type_name)
)
values = []
for key, value in entries.items():
if key not in existing:
values.append(
(type_name, key, json.dumps(value))
)
if values:
await ctx.config.put_many(self.workspace, values)
written += len(values)
ctx.logger.info(
f"Workspace {self.workspace!r} upsert-missing: "
f"{written} new entries"
)