teleo-infrastructure/hermes-agent/leoclean-bin/cloudsql_memory_tool.py
twentyOne2x 8e3d1877c6
Some checks are pending
CI / lint-and-test (push) Waiting to run
Source-control leoclean KB bridge (#59)
2026-07-09 01:09:53 +02:00

1100 lines
44 KiB
Python
Executable file

#!/usr/bin/env python3
"""Cloud SQL memory and KB-proposal bridge for the GCP leoclean profile.
Canonical claim/strategy application remains review-gated. Read commands query
the canonical Teleo KB first, then fall back to the restored `teleo_restore`
shadow schema in Cloud SQL. Proposal commands write staged review records into
`kb_stage.kb_proposals`; they do not apply canonical truth directly.
"""
from __future__ import annotations
import argparse
import json
import os
import re
import subprocess
from typing import Any
STOPWORDS = {
"a", "about", "after", "all", "an", "and", "are", "as", "at", "be", "before",
"by", "can", "cloud", "context", "do", "does", "for", "from", "gcp", "give",
"how", "i", "in", "into", "is", "it", "leo", "me", "memory", "of", "on", "or",
"our", "production", "should", "source", "that", "the", "their", "this", "to",
"using", "what", "when", "where", "which", "who", "why", "with", "you", "your",
}
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--host", default=os.environ.get("TELEO_CLOUDSQL_HOST", "10.61.0.3"))
parser.add_argument("--port", default=os.environ.get("TELEO_CLOUDSQL_PORT", "5432"))
parser.add_argument("--db", default=os.environ.get("TELEO_CLOUDSQL_DB", "teleo_kb"))
parser.add_argument("--canonical-db", default=os.environ.get("TELEO_CANONICAL_CLOUDSQL_DB", "teleo_canonical"))
parser.add_argument("--user", default=os.environ.get("TELEO_CLOUDSQL_USER", "postgres"))
parser.add_argument(
"--password-secret",
default=os.environ.get("TELEO_CLOUDSQL_PASSWORD_SECRET", "gcp-teleo-pgvector-standby-postgres-password"),
)
parser.add_argument("--project", default=os.environ.get("TELEO_GCP_PROJECT", "teleo-501523"))
parser.add_argument("--format", choices=["markdown", "json"], default="markdown")
parser.add_argument(
"--include-excerpts",
"--raw-ok",
dest="include_excerpts",
action="store_true",
default=os.environ.get("TELEO_MEMORY_INCLUDE_EXCERPTS", "") == "1",
help="Include short memory excerpts for agent context. Default proof mode is redacted.",
)
parser.add_argument("--redacted", action="store_true", default=os.environ.get("TELEO_MEMORY_REDACT", "") == "1")
sub = parser.add_subparsers(dest="command", required=True)
search = sub.add_parser("search", help="Search restored Cloud SQL memory rows.")
search.add_argument("query")
search.add_argument("--limit", type=int, default=8)
search.add_argument("--context-limit", type=int, default=argparse.SUPPRESS, help="Accepted for teleo-kb compatibility; ignored.")
search.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
search.add_argument("--redacted", action="store_true", default=argparse.SUPPRESS)
search.add_argument("--include-excerpts", "--raw-ok", dest="include_excerpts", action="store_true", default=argparse.SUPPRESS)
context = sub.add_parser("context", help="Build an agent-ready memory context bundle.")
context.add_argument("query")
context.add_argument("--limit", type=int, default=8)
context.add_argument("--context-limit", type=int, default=argparse.SUPPRESS, help="Accepted for teleo-kb compatibility; ignored.")
context.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
context.add_argument("--redacted", action="store_true", default=argparse.SUPPRESS)
context.add_argument("--include-excerpts", "--raw-ok", dest="include_excerpts", action="store_true", default=argparse.SUPPRESS)
show = sub.add_parser("show", help="Show one canonical claim with evidence and edges.")
show.add_argument("claim_id")
show.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
show.add_argument("--redacted", action="store_true", default=argparse.SUPPRESS)
show.add_argument("--include-excerpts", "--raw-ok", dest="include_excerpts", action="store_true", default=argparse.SUPPRESS)
evidence = sub.add_parser("evidence", help="Show canonical evidence rows for one claim.")
evidence.add_argument("claim_id")
evidence.add_argument("--limit", type=int, default=8)
evidence.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
evidence.add_argument("--redacted", action="store_true", default=argparse.SUPPRESS)
evidence.add_argument("--include-excerpts", "--raw-ok", dest="include_excerpts", action="store_true", default=argparse.SUPPRESS)
edges = sub.add_parser("edges", help="Show canonical graph edges for one claim.")
edges.add_argument("claim_id")
edges.add_argument("--limit", type=int, default=12)
edges.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
edges.add_argument("--redacted", action="store_true", default=argparse.SUPPRESS)
edges.add_argument("--include-excerpts", "--raw-ok", dest="include_excerpts", action="store_true", default=argparse.SUPPRESS)
status = sub.add_parser("status", help="Read back Cloud SQL memory backend status.")
status.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
status.add_argument("--redacted", action="store_true", default=argparse.SUPPRESS)
propose = sub.add_parser(
"propose-core-change",
help="Stage a review-gated canonical KB change instead of writing core truth to runtime memory.",
)
propose.add_argument("--proposal-type", choices=["revise_claim", "revise_strategy"], required=True)
propose.add_argument("--target-kind", choices=["claim", "belief", "strategy", "identity", "role", "telos", "framework", "reasoning_tool", "other"], required=True)
propose.add_argument("--target-ref", required=True, help="Claim id, strategy id/name, or unresolved target description.")
propose.add_argument("--current", default="", help="Current/old claim or state, if known.")
propose.add_argument("--proposed", required=True, help="Proposed replacement or correction.")
propose.add_argument("--evidence", action="append", default=[], help="Evidence/source pointer. Can be repeated.")
propose.add_argument("--implication", action="append", default=[], help="Downstream implication. Can be repeated.")
propose.add_argument("--proposed-by", default="leo")
propose.add_argument("--originator", default="", help="Human or agent who originated the correction.")
propose.add_argument("--channel", default="telegram")
propose.add_argument("--source-ref", required=True)
propose.add_argument("--rationale", required=True)
propose.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
list_proposals = sub.add_parser("list-proposals", help="List staged KB proposals from Cloud SQL.")
list_proposals.add_argument("--status", default="pending_review")
list_proposals.add_argument("--limit", type=int, default=10)
list_proposals.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
show_proposal = sub.add_parser("show-proposal", help="Show one staged KB proposal from Cloud SQL.")
show_proposal.add_argument("proposal_id")
show_proposal.add_argument("--format", choices=["markdown", "json"], default=argparse.SUPPRESS)
return parser.parse_args()
def terms_for(query: str) -> list[str]:
terms: list[str] = []
for token in re.findall(r"[A-Za-z0-9][A-Za-z0-9.+#/-]*", query.lower()):
token = token.strip("-/")
if len(token) < 3 or token in STOPWORDS:
continue
if token not in terms:
terms.append(token)
return terms[:14] or [query.lower()]
def sql_literal(value: str) -> str:
return "'" + value.replace("'", "''") + "'"
def sql_json_array(values: list[str]) -> str:
return "'{}'::json".format(json.dumps(values).replace("'", "''"))
def sql_array(values: list[str], cast: str = "text") -> str:
return "array[" + ",".join(sql_literal(value) for value in values) + f"]::{cast}[]"
def password(args: argparse.Namespace) -> str:
if os.environ.get("PGPASSWORD"):
return os.environ["PGPASSWORD"]
result = subprocess.run(
[
"gcloud",
"secrets",
"versions",
"access",
"latest",
f"--secret={args.password_secret}",
f"--project={args.project}",
],
text=True,
capture_output=True,
check=False,
)
if result.returncode != 0:
raise SystemExit(f"Secret access failed: {result.stderr.strip()}")
return result.stdout.strip()
def run_psql(args: argparse.Namespace, sql: str, db: str | None = None) -> str:
env = os.environ.copy()
env["PGPASSWORD"] = password(args)
conn = f"host={args.host} port={args.port} dbname={db or args.db} user={args.user} sslmode=require"
result = subprocess.run(
["psql", conn, "-At", "-q", "-v", "ON_ERROR_STOP=1"],
text=True,
capture_output=True,
input=sql,
env=env,
check=False,
)
if result.returncode != 0:
raise SystemExit(
f"Cloud SQL query failed ({result.returncode})\nSTDOUT:\n{result.stdout}\nSTDERR:\n{result.stderr}"
)
return result.stdout
def psql_json_lines(args: argparse.Namespace, sql: str, db: str | None = None) -> list[dict[str, Any]]:
rows: list[dict[str, Any]] = []
for line in run_psql(args, sql, db=db).splitlines():
line = line.strip()
if line:
rows.append(json.loads(line))
return rows
def include_excerpts(args: argparse.Namespace) -> bool:
return bool(getattr(args, "include_excerpts", False)) and not bool(getattr(args, "redacted", False))
def query_canonical_rows(args: argparse.Namespace, query: str, limit: int) -> dict[str, Any]:
terms = terms_for(query)
patterns = [f"%{term}%" for term in terms]
min_score = 2 if len(patterns) >= 3 else 1
claim_limit = max(1, min(limit, 20))
context_limit = max(1, min(int(getattr(args, "context_limit", limit) or limit), 20))
claims_sql = f"""
with params(patterns) as (
values ({sql_array(patterns)})
), ranked as (
select c.id,
c.type,
c.text,
c.status,
c.confidence,
c.tags,
(
select count(*)
from params p, unnest(p.patterns) pattern
where lower(c.text) like pattern
or exists (
select 1 from unnest(coalesce(c.tags, '{{}}'::text[])) tag
where lower(tag) like pattern
)
) as score,
(select count(*) from public.claim_evidence ce where ce.claim_id = c.id) as evidence_count,
(select count(*) from public.claim_edges e where e.from_claim = c.id or e.to_claim = c.id) as edge_count
from public.claims c
where c.status = 'open'
)
select jsonb_build_object(
'id', id::text,
'type', type,
'text', text,
'status', status,
'confidence', confidence,
'tags', coalesce(to_jsonb(tags), '[]'::jsonb),
'score', score,
'evidence_count', evidence_count,
'edge_count', edge_count
)::text
from ranked
where score >= {min_score}
order by score desc, evidence_count desc, edge_count desc, coalesce(confidence, 0) desc, length(text)
limit {claim_limit};
"""
context_sql = f"""
with params(patterns) as (
values ({sql_array(patterns)})
), corpus as (
select 'persona' as source,
a.handle as owner,
p.name as title,
concat_ws(E'\\n', p.voice, p.role, p.source_ref, p.lens) as body
from public.personas p
join public.agents a on a.id = p.agent_id
union all
select 'strategy' as source,
a.handle as owner,
'active strategy v' || s.version::text as title,
concat_ws(E'\\n', s.diagnosis, s.guiding_policy, s.proximate_objectives::text) as body
from public.strategies s
join public.agents a on a.id = s.agent_id
where s.active
union all
select 'belief' as source,
a.handle as owner,
concat_ws(' ', b.level::text, 'rank', b.rank::text) as title,
concat_ws(E'\\n', b.statement, b.falsifier) as body
from public.beliefs b
join public.agents a on a.id = b.agent_id
where b.status = 'active'
union all
select 'blindspot' as source,
a.handle as owner,
bs.name as title,
concat_ws(E'\\n', bs.description, bs.correction, bs.kind) as body
from public.blindspots bs
join public.agents a on a.id = bs.agent_id
where bs.status = 'active'
union all
select 'agent_role' as source,
a.handle as owner,
ar.title as title,
ar.description as body
from public.agent_roles ar
join public.agents a on a.id = ar.agent_id
union all
select 'peer_model' as source,
subject.handle as owner,
'peer ' || peer.handle || ': ' || pm.domain as title,
concat_ws(E'\\n', pm.outranks_on, pm.deference_rule) as body
from public.peer_models pm
join public.agents subject on subject.id = pm.subject_id
join public.agents peer on peer.id = pm.peer_id
union all
select 'behavioral_rule' as source,
coalesce(a.handle, 'collective') as owner,
br.category::text as title,
concat_ws(E'\\n', br.rule, br.rationale) as body
from public.behavioral_rules br
left join public.agents a on a.id = br.agent_id
union all
select 'contributor_rule' as source,
coalesce(a.handle, 'collective') as owner,
cr.name as title,
concat_ws(E'\\n', cr.directive, cr.ci_tier, cr.weighting, cr.rationale) as body
from public.contributor_rules cr
left join public.agents a on a.id = cr.agent_id
union all
select 'reasoning_tool' as source,
coalesce(a.handle, 'collective') as owner,
rt.name as title,
concat_ws(E'\\n', rt.description, rt.category) as body
from public.reasoning_tools rt
left join public.agents a on a.id = rt.agent_id
union all
select 'governance_gate' as source,
coalesce(a.handle, 'collective') as owner,
gg.name as title,
concat_ws(E'\\n', gg.criteria, gg.evidence_bar, gg.pass_condition) as body
from public.governance_gates gg
left join public.agents a on a.id = gg.agent_id
), ranked as (
select source,
owner,
title,
body,
(
select count(*)
from params p, unnest(p.patterns) pattern
where lower(concat_ws(E'\\n', source, owner, title, body)) like pattern
) as score
from corpus
)
select jsonb_build_object(
'source', source,
'owner', owner,
'title', title,
'body', left(coalesce(body, ''), 900),
'score', score
)::text
from ranked
where score >= {min_score}
order by score desc, source, owner, title
limit {context_limit};
"""
claims = psql_json_lines(args, claims_sql, db=args.canonical_db)
claim_ids = [claim["id"] for claim in claims]
evidence: dict[str, list[dict[str, Any]]] = {claim_id: [] for claim_id in claim_ids}
edges: dict[str, list[dict[str, Any]]] = {claim_id: [] for claim_id in claim_ids}
if claim_ids:
ids = sql_array(claim_ids, "uuid")
evidence_sql = f"""
with ranked as (
select ce.claim_id,
ce.role::text as role,
ce.weight,
s.source_type,
s.url,
s.storage_path,
left(coalesce(s.excerpt, ''), 800) as excerpt,
row_number() over (
partition by ce.claim_id
order by (s.storage_path like 'inbox/archive/%') desc,
ce.role::text,
s.storage_path nulls last,
s.url nulls last
) as rn
from public.claim_evidence ce
join public.sources s on s.id = ce.source_id
where ce.claim_id = any({ids})
)
select jsonb_build_object(
'claim_id', claim_id::text,
'role', role,
'weight', weight,
'source_type', source_type,
'url', url,
'storage_path', storage_path,
'excerpt', excerpt
)::text
from ranked
where rn <= 4
order by claim_id, rn;
"""
for row in psql_json_lines(args, evidence_sql, db=args.canonical_db):
evidence.setdefault(row["claim_id"], []).append(row)
edges_sql = f"""
with base(id) as (
select unnest({ids})
), edge_rows as (
select base.id as claim_id,
case when e.from_claim = base.id then 'outgoing' else 'incoming' end as direction,
e.edge_type::text as edge_type,
other.id as connected_id,
other.text as connected_text,
other.tags as connected_tags,
row_number() over (
partition by base.id
order by case e.edge_type::text
when 'supports' then 1
when 'challenges' then 2
when 'contradicts' then 3
when 'requires' then 4
when 'supersedes' then 5
when 'constrains' then 6
when 'causes' then 7
when 'accelerates' then 8
else 9
end,
other.text
) as rn
from base
join public.claim_edges e on e.from_claim = base.id or e.to_claim = base.id
join public.claims other on other.id = case when e.from_claim = base.id then e.to_claim else e.from_claim end
)
select jsonb_build_object(
'claim_id', claim_id::text,
'direction', direction,
'edge_type', edge_type,
'connected_id', connected_id::text,
'connected_text', connected_text,
'connected_tags', coalesce(to_jsonb(connected_tags), '[]'::jsonb)
)::text
from edge_rows
where rn <= 6
order by claim_id, rn;
"""
for row in psql_json_lines(args, edges_sql, db=args.canonical_db):
edges.setdefault(row["claim_id"], []).append(row)
for claim in claims:
claim["evidence"] = evidence.get(claim["id"], [])
claim["edges"] = edges.get(claim["id"], [])
context_rows = psql_json_lines(args, context_sql, db=args.canonical_db)
result: dict[str, Any] = {
"artifact": "teleo_cloudsql_canonical_kb_context",
"backend": f"cloudsql:teleo-pgvector-standby/{args.canonical_db}/public+kb_stage",
"query": query,
"terms": terms,
"claims": claims,
"context_rows": context_rows,
"hit_count_total": len(claims) + len(context_rows),
"hits": [
{
"source_table": "public.claims",
"row_id": claim["id"],
"score": claim.get("score"),
"claim_type": claim.get("type"),
"claim_status": claim.get("status"),
"confidence": claim.get("confidence"),
"tags": claim.get("tags") or [],
"evidence_count": claim.get("evidence_count"),
"edge_count": claim.get("edge_count"),
"excerpt": claim.get("text"),
}
for claim in claims
] + [
{
"source_table": f"public.{row['source']}",
"row_id": f"{row['owner']}:{row['title']}",
"score": row.get("score"),
"actor": row.get("owner"),
"excerpt": row.get("body"),
}
for row in context_rows
],
"privacy": "redacted proof mode by default; pass --include-excerpts/--raw-ok only for private agent context",
}
if not include_excerpts(args):
for claim in result["claims"]:
claim.pop("text", None)
for row in claim.get("evidence", []):
row.pop("excerpt", None)
for row in claim.get("edges", []):
row.pop("connected_text", None)
for row in result["context_rows"]:
row.pop("body", None)
for hit in result["hits"]:
hit.pop("excerpt", None)
result["privacy"] = "redacted proof mode: raw body used only inside SQL matching; output has counts, hashes, canonical table, row id, timestamps, and evidence links"
else:
result["privacy"] = "private agent context mode: short excerpts included; do not retain or paste raw output into public artifacts"
return result
def query_audit_rows(args: argparse.Namespace, query: str, limit: int) -> dict[str, Any]:
terms = terms_for(query)
sql = f"""
with q(term) as (
select lower(value) from json_array_elements_text({sql_json_array(terms)}) as value
), raw_hits as (
select
'response_audit'::text as source_table,
id::text as row_id,
coalesce(timestamp, created_at) as event_time,
coalesce("user", agent, '') as actor,
concat_ws(' ', query, conversation_window, entities_matched, claims_matched,
retrieval_layers_hit, retrieval_gap, market_data, research_context, kb_context_text,
tool_calls, raw_response, display_response, query_type) as body
from teleo_restore.response_audit
union all
select
'sources'::text as source_table,
path as row_id,
coalesce(updated_at, created_at) as event_time,
coalesce(submitted_by, original_author, original_author_handle, '') as actor,
concat_ws(' ', path, status, priority, priority_log, extraction_model, last_error,
feedback, content_type, original_author, original_author_handle) as body
from teleo_restore.sources
union all
select
'agent_research_runs'::text as source_table,
id as row_id,
coalesce(completed_at, created_at) as event_time,
coalesce(agent_slug, '') as actor,
concat_ws(' ', source_surface, source_ref, request_kind, prompt_excerpt,
selected_provider, selected_route, status, answer_excerpt, proof_ref) as body
from teleo_restore.agent_research_runs
union all
select
'agent_tool_invocations'::text as source_table,
id as row_id,
created_at as event_time,
coalesce(provider, tool_name, '') as actor,
concat_ws(' ', provider, tool_name, tool_category, endpoint_host, decision,
decision_reason, rail, network, currency, error_class) as body
from teleo_restore.agent_tool_invocations
), matched as (
select *,
(select count(*) from q where lower(raw_hits.body) like '%' || q.term || '%') as score
from raw_hits
where exists (select 1 from q where lower(raw_hits.body) like '%' || q.term || '%')
), ranked as (
select *, row_number() over (order by score desc, event_time desc nulls last, source_table, row_id) as rn
from matched
)
select json_build_object(
'hit_count_total', (select count(*) from matched),
'hits', coalesce((select json_agg(json_build_object(
'source_table', source_table,
'row_id', row_id,
'event_time', event_time,
'actor', actor,
'score', score,
'body_chars', length(body),
'body_md5', md5(body),
'excerpt', left(regexp_replace(body, '\\s+', ' ', 'g'), 700)
) order by rn) from ranked where rn <= {max(1, min(limit, 20))}), '[]'::json)
)::text;
"""
text = run_psql(args, sql).strip()
result = json.loads(text)
result.update(
{
"artifact": "teleo_cloudsql_memory_context",
"backend": "cloudsql:teleo-pgvector-standby/teleo_kb/teleo_restore",
"query": query,
"terms": terms,
"privacy": "redacted proof mode by default; pass --include-excerpts/--raw-ok only for private agent context",
}
)
if not include_excerpts(args):
for hit in result["hits"]:
hit.pop("excerpt", None)
result["privacy"] = "redacted proof mode: raw body used only inside SQL matching; output has counts, hashes, source table, row id, and timestamps"
else:
result["privacy"] = "private agent context mode: short excerpts included; do not retain or paste raw output into public artifacts"
return result
def query_rows(args: argparse.Namespace, query: str, limit: int) -> dict[str, Any]:
canonical = query_canonical_rows(args, query, limit)
if canonical.get("hit_count_total", 0) or canonical.get("hits"):
return canonical
audit = query_audit_rows(args, query, limit)
audit["canonical_fallback_reason"] = "no matching canonical public/persona/strategy/belief rows"
return audit
def canonical_backend(args: argparse.Namespace) -> str:
return f"cloudsql:teleo-pgvector-standby/{args.canonical_db}/public+kb_stage"
def canonical_claim(args: argparse.Namespace, claim_id: str) -> dict[str, Any] | None:
sql = f"""
select jsonb_build_object(
'id', id::text,
'type', type,
'text', text,
'status', status,
'confidence', confidence,
'tags', coalesce(to_jsonb(tags), '[]'::jsonb),
'superseded_by', superseded_by::text,
'created_at', created_at::text,
'updated_at', updated_at::text
)::text
from public.claims
where id = {sql_literal(claim_id)}::uuid;
"""
rows = psql_json_lines(args, sql, db=args.canonical_db)
return rows[0] if rows else None
def canonical_evidence(args: argparse.Namespace, claim_id: str, limit: int) -> list[dict[str, Any]]:
sql = f"""
with ranked as (
select ce.claim_id,
ce.role::text as role,
ce.weight,
s.source_type,
s.url,
s.storage_path,
left(coalesce(s.excerpt, ''), 800) as excerpt,
row_number() over (
partition by ce.claim_id
order by (s.storage_path like 'inbox/archive/%') desc,
ce.role::text,
s.storage_path nulls last,
s.url nulls last
) as rn
from public.claim_evidence ce
join public.sources s on s.id = ce.source_id
where ce.claim_id = {sql_literal(claim_id)}::uuid
)
select jsonb_build_object(
'claim_id', claim_id::text,
'role', role,
'weight', weight,
'source_type', source_type,
'url', url,
'storage_path', storage_path,
'excerpt', excerpt
)::text
from ranked
where rn <= {max(1, min(limit, 50))}
order by rn;
"""
rows = psql_json_lines(args, sql, db=args.canonical_db)
if not include_excerpts(args):
for row in rows:
row.pop("excerpt", None)
return rows
def canonical_edges(args: argparse.Namespace, claim_id: str, limit: int) -> list[dict[str, Any]]:
sql = f"""
with base(id) as (
values ({sql_literal(claim_id)}::uuid)
), edge_rows as (
select base.id as claim_id,
case when e.from_claim = base.id then 'outgoing' else 'incoming' end as direction,
e.edge_type::text as edge_type,
other.id as connected_id,
other.text as connected_text,
other.tags as connected_tags,
row_number() over (
partition by base.id
order by case e.edge_type::text
when 'supports' then 1
when 'challenges' then 2
when 'contradicts' then 3
when 'requires' then 4
when 'supersedes' then 5
when 'constrains' then 6
when 'causes' then 7
when 'accelerates' then 8
else 9
end,
other.text
) as rn
from base
join public.claim_edges e on e.from_claim = base.id or e.to_claim = base.id
join public.claims other on other.id = case when e.from_claim = base.id then e.to_claim else e.from_claim end
)
select jsonb_build_object(
'claim_id', claim_id::text,
'direction', direction,
'edge_type', edge_type,
'connected_id', connected_id::text,
'connected_text', connected_text,
'connected_tags', coalesce(to_jsonb(connected_tags), '[]'::jsonb)
)::text
from edge_rows
where rn <= {max(1, min(limit, 50))}
order by rn;
"""
rows = psql_json_lines(args, sql, db=args.canonical_db)
if not include_excerpts(args):
for row in rows:
row.pop("connected_text", None)
return rows
def show_canonical_claim(args: argparse.Namespace) -> dict[str, Any]:
claim = canonical_claim(args, args.claim_id)
if not claim:
raise SystemExit(f"claim not found: {args.claim_id}")
if not include_excerpts(args):
claim.pop("text", None)
return {
"artifact": "teleo_cloudsql_canonical_kb_claim",
"backend": canonical_backend(args),
"claim": claim,
"evidence": canonical_evidence(args, args.claim_id, 8),
"edges": canonical_edges(args, args.claim_id, 12),
}
def evidence_canonical_claim(args: argparse.Namespace) -> dict[str, Any]:
return {
"artifact": "teleo_cloudsql_canonical_kb_evidence",
"backend": canonical_backend(args),
"claim_id": args.claim_id,
"evidence": canonical_evidence(args, args.claim_id, args.limit),
}
def edges_canonical_claim(args: argparse.Namespace) -> dict[str, Any]:
return {
"artifact": "teleo_cloudsql_canonical_kb_edges",
"backend": canonical_backend(args),
"claim_id": args.claim_id,
"edges": canonical_edges(args, args.claim_id, args.limit),
}
def propose_core_change(args: argparse.Namespace) -> dict[str, Any]:
payload = {
"target_kind": args.target_kind,
"target_ref": args.target_ref,
"current": args.current or None,
"proposed": args.proposed,
"evidence": args.evidence,
"implications": args.implication,
"originator": args.originator or None,
"routing_rule": "core Teleo changes must be staged in kb_stage.kb_proposals, not saved as runtime memory",
"canonical_apply_required": True,
}
payload_json = json.dumps(payload, sort_keys=True)
sql = f"""
with params as (
select
{sql_literal(args.proposal_type)}::text as proposal_type,
nullif({sql_literal(args.proposed_by)}, '') as proposed_by_handle,
nullif({sql_literal(args.channel)}, '') as channel,
nullif({sql_literal(args.source_ref)}, '') as source_ref,
{sql_literal(args.rationale)} as rationale,
{sql_literal(payload_json)}::jsonb as payload
), proposer as (
select a.id, a.handle
from public.agents a, params p
where a.handle = p.proposed_by_handle
limit 1
), inserted as (
insert into kb_stage.kb_proposals (
proposal_type,
status,
proposed_by_handle,
proposed_by_agent_id,
channel,
source_ref,
rationale,
payload
)
select
p.proposal_type,
'pending_review',
p.proposed_by_handle,
proposer.id,
coalesce(p.channel, 'telegram'),
p.source_ref,
p.rationale,
p.payload
from params p
left join proposer on true
returning *
)
select jsonb_build_object(
'artifact', 'teleo_cloudsql_kb_core_change_proposal',
'backend', {sql_literal(canonical_backend(args))},
'id', id::text,
'proposal_type', proposal_type,
'status', status,
'proposed_by_handle', proposed_by_handle,
'proposed_by_agent_id', proposed_by_agent_id::text,
'channel', channel,
'source_ref', source_ref,
'rationale', rationale,
'payload', payload,
'created_at', created_at::text,
'canonical_apply_done', false,
'runtime_memory_write_done', false
)::text
from inserted;
"""
rows = psql_json_lines(args, sql, db=args.canonical_db)
if not rows:
raise SystemExit("No core-change proposal inserted.")
return rows[0]
def list_proposals(args: argparse.Namespace) -> dict[str, Any]:
where = ""
if args.status.lower() != "all":
where = f"where status = {sql_literal(args.status)}"
sql = f"""
select jsonb_build_object(
'id', id::text,
'proposal_type', proposal_type,
'status', status,
'proposed_by_handle', proposed_by_handle,
'channel', channel,
'source_ref', source_ref,
'rationale', left(rationale, 500),
'payload', payload,
'created_at', created_at::text,
'reviewed_at', reviewed_at::text,
'applied_at', applied_at::text
)::text
from kb_stage.kb_proposals
{where}
order by created_at desc
limit {max(1, min(args.limit, 100))};
"""
return {
"artifact": "teleo_cloudsql_kb_proposal_list",
"backend": canonical_backend(args),
"status_filter": args.status,
"proposals": psql_json_lines(args, sql, db=args.canonical_db),
}
def show_proposal(args: argparse.Namespace) -> dict[str, Any]:
sql = f"""
select jsonb_build_object(
'artifact', 'teleo_cloudsql_kb_proposal',
'backend', {sql_literal(canonical_backend(args))},
'id', id::text,
'proposal_type', proposal_type,
'status', status,
'proposed_by_handle', proposed_by_handle,
'proposed_by_agent_id', proposed_by_agent_id::text,
'channel', channel,
'source_ref', source_ref,
'rationale', rationale,
'payload', payload,
'reviewed_by_handle', reviewed_by_handle,
'reviewed_at', reviewed_at::text,
'review_note', review_note,
'applied_by_handle', applied_by_handle,
'applied_at', applied_at::text,
'created_at', created_at::text,
'updated_at', updated_at::text
)::text
from kb_stage.kb_proposals
where id = {sql_literal(args.proposal_id)}::uuid;
"""
rows = psql_json_lines(args, sql, db=args.canonical_db)
if not rows:
raise SystemExit(f"proposal not found: {args.proposal_id}")
return rows[0]
def status(args: argparse.Namespace) -> dict[str, Any]:
canonical_sql = """
select json_build_object(
'db_identity', current_database() || '|' || current_user,
'extensions', (select json_agg(extname order by extname) from pg_extension),
'schema_tables', (
select json_object_agg(table_schema, table_count) from (
select table_schema, count(*) as table_count
from information_schema.tables
where table_type='BASE TABLE' and table_schema in ('public','kb_stage')
group by table_schema
order by table_schema
) counts
),
'high_signal_rows', json_build_object(
'agents', (select count(*) from public.agents),
'personas', (select count(*) from public.personas),
'strategies', (select count(*) from public.strategies),
'beliefs', (select count(*) from public.beliefs),
'claims', (select count(*) from public.claims),
'claim_edges', (select count(*) from public.claim_edges),
'claim_evidence', (select count(*) from public.claim_evidence),
'sources', (select count(*) from public.sources),
'kb_proposals', (select count(*) from kb_stage.kb_proposals)
)
)::text;
"""
audit_sql = """
select json_build_object(
'db_identity', current_database() || '|' || current_user,
'extensions', (select json_agg(extname order by extname) from pg_extension),
'teleo_restore_tables', (select count(*) from pg_tables where schemaname='teleo_restore'),
'total_rows', (
select sum(row_count) from (
select count(*)::bigint as row_count from teleo_restore.response_audit
union all select count(*)::bigint from teleo_restore.sources
union all select count(*)::bigint from teleo_restore.agent_research_runs
union all select count(*)::bigint from teleo_restore.agent_tool_invocations
union all select count(*)::bigint from teleo_restore.audit_log
union all select count(*)::bigint from teleo_restore.metrics_snapshots
union all select count(*)::bigint from teleo_restore.prs
union all select count(*)::bigint from teleo_restore.review_records
) counts
),
'high_signal_rows', json_build_object(
'response_audit', (select count(*) from teleo_restore.response_audit),
'sources', (select count(*) from teleo_restore.sources),
'agent_research_runs', (select count(*) from teleo_restore.agent_research_runs),
'agent_tool_invocations', (select count(*) from teleo_restore.agent_tool_invocations)
)
)::text;
"""
result = json.loads(run_psql(args, audit_sql).strip())
result["canonical"] = json.loads(run_psql(args, canonical_sql, db=args.canonical_db).strip())
result["artifact"] = "teleo_cloudsql_memory_status"
result["backend"] = f"cloudsql:teleo-pgvector-standby/{args.canonical_db}/public+kb_stage primary; {args.db}/teleo_restore fallback"
return result
def emit_json(value: dict[str, Any]) -> None:
print(json.dumps(value, indent=2))
def emit_markdown(value: dict[str, Any]) -> None:
if value["artifact"] == "teleo_cloudsql_memory_status":
print("# Teleo Cloud SQL Memory Status\n")
print(f"- Backend: `{value['backend']}`")
print(f"- DB: `{value['db_identity']}`")
if "canonical" in value:
canonical = value["canonical"]
print(f"- Canonical DB: `{canonical['db_identity']}`")
print(f"- Canonical tables: `{json.dumps(canonical.get('schema_tables'), sort_keys=True)}`")
print(f"- Canonical high-signal rows: `{json.dumps(canonical.get('high_signal_rows'), sort_keys=True)}`")
print(f"- Extensions: {', '.join(value.get('extensions') or [])}")
print(f"- Restored tables: {value['teleo_restore_tables']}")
print(f"- Approx restored rows checked: {value['total_rows']}")
print(f"- High-signal rows: `{json.dumps(value['high_signal_rows'], sort_keys=True)}`")
return
if "claims" in value:
print("# Teleo KB Context\n")
print(f"- Backend: `{value['backend']}`")
print(f"- Query: `{value['query']}`")
print(f"- Terms: {', '.join(value['terms'])}")
print(f"- Privacy: {value['privacy']}\n")
if value.get("context_rows"):
print("## Soul / Context Rows\n")
for row in value["context_rows"]:
print(
f"- `{row.get('source')}` / `{row.get('owner')}` / `{row.get('title')}` "
f"(score {row.get('score')}): {row.get('body', '[redacted]')}"
)
print()
print("## Claims\n")
if not value.get("claims"):
print("No matching canonical claims found.")
return
for i, claim in enumerate(value["claims"], 1):
print(f"### {i}. {claim.get('text', '[redacted claim text]')}\n")
print(f"- id: `{claim['id']}`")
print(f"- type: `{claim.get('type')}`; confidence: `{claim.get('confidence')}`; score: `{claim.get('score')}`")
print(f"- tags: `{', '.join(claim.get('tags') or [])}`")
print(f"- evidence rows: `{claim.get('evidence_count', len(claim.get('evidence', [])))}`")
print(f"- edge rows: `{claim.get('edge_count', len(claim.get('edges', [])))}`")
if claim.get("evidence"):
print("\nEvidence:")
for ev in claim["evidence"]:
source = ev.get("storage_path") or ev.get("url") or "(no source pointer)"
excerpt = f" - {ev['excerpt']}" if ev.get("excerpt") else ""
print(f"- `{ev.get('role')}` / `{ev.get('source_type')}` / `{source}`{excerpt}")
if claim.get("edges"):
print("\nEdges:")
for edge in claim["edges"]:
text = f": {edge['connected_text']}" if edge.get("connected_text") else ""
print(f"- `{edge.get('direction')}` `{edge.get('edge_type')}` ({edge.get('connected_id')}){text}")
print()
return
if value["artifact"] == "teleo_cloudsql_canonical_kb_claim":
claim = value["claim"]
print("# Teleo KB Claim\n")
print(f"- Backend: `{value['backend']}`")
print(f"- id: `{claim['id']}`")
print(f"- type: `{claim.get('type')}`; status: `{claim.get('status')}`; confidence: `{claim.get('confidence')}`")
print(f"- tags: `{', '.join(claim.get('tags') or [])}`")
if claim.get("text"):
print(f"\n{claim['text']}\n")
print("## Evidence\n")
for ev in value.get("evidence", []):
source = ev.get("storage_path") or ev.get("url") or "(no source pointer)"
excerpt = f" - {ev['excerpt']}" if ev.get("excerpt") else ""
print(f"- `{ev.get('role')}` / `{ev.get('source_type')}` / `{source}`{excerpt}")
print("\n## Edges\n")
for edge in value.get("edges", []):
text = f": {edge['connected_text']}" if edge.get("connected_text") else ""
print(f"- `{edge.get('direction')}` `{edge.get('edge_type')}` ({edge.get('connected_id')}){text}")
return
if value["artifact"] in {"teleo_cloudsql_canonical_kb_evidence", "teleo_cloudsql_canonical_kb_edges"}:
print(f"# {value['artifact']}\n")
print(f"- Backend: `{value['backend']}`")
print(f"- Claim: `{value['claim_id']}`\n")
rows = value.get("evidence") or value.get("edges") or []
for row in rows:
print(f"- `{json.dumps(row, sort_keys=True)}`")
return
if value["artifact"] == "teleo_cloudsql_kb_core_change_proposal":
print("# Staged KB Core-Change Proposal\n")
print(f"- Proposal id: `{value['id']}`")
print(f"- Type: `{value['proposal_type']}`")
print(f"- Status: `{value['status']}`")
print(f"- Source: `{value.get('source_ref') or '-'}`")
print(f"- Runtime memory write: `{value['runtime_memory_write_done']}`")
print(f"- Canonical apply done: `{value['canonical_apply_done']}`")
print("\n## Rationale\n")
print(value["rationale"])
print("\n## Payload\n")
print(json.dumps(value.get("payload") or {}, indent=2, sort_keys=True))
return
if value["artifact"] == "teleo_cloudsql_kb_proposal_list":
print("# Staged KB Proposals\n")
rows = value.get("proposals") or []
if not rows:
print("No matching proposals.")
for row in rows:
print(f"## {row['id']}\n")
print(f"- Type: `{row['proposal_type']}`")
print(f"- Status: `{row['status']}`")
print(f"- Source: `{row.get('source_ref') or '-'}`")
print(f"- Created: `{row.get('created_at')}`")
print(f"- Rationale: {row.get('rationale') or ''}")
print()
return
if value["artifact"] == "teleo_cloudsql_kb_proposal":
print("# Staged KB Proposal\n")
print(f"- Proposal id: `{value['id']}`")
print(f"- Type: `{value['proposal_type']}`")
print(f"- Status: `{value['status']}`")
print(f"- Source: `{value.get('source_ref') or '-'}`")
print(f"- Created: `{value.get('created_at')}`")
print(f"- Applied at: `{value.get('applied_at') or '-'}`")
print("\n## Rationale\n")
print(value["rationale"])
print("\n## Payload\n")
print(json.dumps(value.get("payload") or {}, indent=2, sort_keys=True))
return
print("# Teleo Cloud SQL Memory Context\n")
print(f"- Backend: `{value['backend']}`")
print(f"- Query: `{value['query']}`")
print(f"- Terms: {', '.join(value['terms'])}")
print(f"- Total matching rows: {value['hit_count_total']}")
print(f"- Privacy: {value['privacy']}\n")
for i, hit in enumerate(value["hits"], 1):
print(f"## Hit {i}: {hit['source_table']} / {hit['row_id']}")
print(f"- Time: {hit.get('event_time') or 'unknown'}")
print(f"- Actor/source: {hit.get('actor') or 'unknown'}")
print(f"- Score: {hit.get('score')}")
print(f"- Body chars: {hit.get('body_chars')}")
print(f"- Body md5: `{hit.get('body_md5')}`")
if "excerpt" in hit:
print(f"- Excerpt: {hit['excerpt']}")
print()
def main() -> None:
args = parse_args()
if args.command == "status":
result = status(args)
elif args.command == "show":
result = show_canonical_claim(args)
elif args.command == "evidence":
result = evidence_canonical_claim(args)
elif args.command == "edges":
result = edges_canonical_claim(args)
elif args.command == "propose-core-change":
result = propose_core_change(args)
elif args.command == "list-proposals":
result = list_proposals(args)
elif args.command == "show-proposal":
result = show_proposal(args)
else:
result = query_rows(args, args.query, args.limit)
if args.format == "json":
emit_json(result)
else:
emit_markdown(result)
if __name__ == "__main__":
main()