teicor-sdk 0.2.8

Python SDK for Teicor external app integrations

Teicor Python SDK

For full external integration setup (install flow, auth, core/runtime interaction, and troubleshooting), see:

• SDK hub
• External App Integration Guide
• AI Agent Integration Spec

This SDK helps installed Teicor apps interact with:

• Runtime proxy APIs
• Core app-scoped state in Teicor core DB

Install from PyPI

Add to your requirements.txt (pin versions):

    teicor-sdk==0.2.8

Example install command from PyPI:

pip install teicor-sdk==0.2.8

Usage

from teicor_sdk import TeicorClient

client = TeicorClient(
    base_url="https://api.teicor.com",
    team_slug="client-a",
    app_slug="shopify",
    service_token="tkmkt_...",
)

context = client.get_context()
state = client.get_core_state()
state = client.patch_core_state({"cursor": "next-page-token"})
rows = client.runtime_request(method="GET", runtime_path="api/schemas")

runtime_request (and CRUD helpers that call it) now auto-route by mode:

• proxy mode: calls Teicor proxy (/v1/teams/<team_slug>/runtime/...).
• direct mode: fetches/refreshes authority token and calls runtime_api_base_url directly.
• If direct call fails with network/timeout or fallback 5xx statuses, SDK retries once via proxy fallback guidance.

SDK CRUD quickstart (schema/table/column/record)

from teicor_sdk import TeicorClient

client = TeicorClient(
    base_url="https://api.teicor.com",
    team_slug="client-a",
    app_slug="shopify",
    service_token="tkmkt_...",
)

# Context + runtime access descriptor
ctx = client.get_context()
print(ctx.runtime_access_mode)      # proxy (today), direct for eligible installs
print(ctx.runtime_proxy_base_url)   # always available
print(ctx.runtime_api_base_url)     # used for direct mode
print(ctx.runtime_fallback)         # deterministic proxy fallback guidance

if ctx.runtime_access_mode == "direct":
    authority = client.get_runtime_authority_token(scope="write")
    print(authority.runtime_authority_token)
    print(authority.expires_in_seconds)

# ---------- Schema ----------
client.create_schema("app_shopify")
schemas = client.list_schemas()
client.update_schema("app_shopify", new_schema_name="app_shopify_v2")

# ---------- Table ----------
table = client.create_table(
    name="oauth_tokens",
    schema_name="app_shopify_v2",
    is_private=True,
)
table_id = str(table["id"])

tables = client.list_tables()
table_details = client.get_table(table_id)
client.update_table(table_id, name="oauth_tokens_v2")

# ---------- Column ----------
column = client.create_column(
    table_id=table_id,
    name="Provider Account Id",
    column_type="single_line_text",  # see TeicorClient.COLUMN_TYPES
    config={"unique": True},
)
column_id = str(column["id"])

columns = client.list_columns(table_id)
client.update_column(
    table_id=table_id,
    column_id=column_id,
    name="Provider Account ID",
    config={"unique": True},
)

# ---------- Record ----------
created = client.create_record(
    table_id=table_id,
    data={column_id: "acct_123"},
)
record_id = str(created["id"])

records = client.list_records(
    table_id=table_id,
    limit=50,
    offset=0,
    filter_expr={
        "and": [
            {
                "column_id": column_id,
                "op": "contains",  # see TeicorClient.FILTER_OPS
                "value": "acct_",
            }
        ]
    },
    sort=[{"column_id": column_id, "direction": "asc"}],
)

first = client.find_record(
    table_id=table_id,
    filter_expr={
        "and": [
            {"column_id": column_id, "op": "eq", "value": "acct_123"}
        ]
    },
)

client.update_record(
    table_id=table_id,
    record_id=record_id,
    data={column_id: "acct_456"},
)

client.delete_record(table_id=table_id, record_id=record_id)

# ---------- Cleanup ----------
client.delete_column(table_id=table_id, column_id=column_id)
client.delete_table(table_id)
client.delete_schema("app_shopify_v2")

Query APIs in SDK

The SDK now exposes query helpers for external integrations:

• list_queries(group_id=None, include_all=False, updated_since=None)
• get_query(query_id)
• execute_query(query_id)
• get_query_data(query_id, if_newer_than_watermark=None)
• sync_queries(...) convenience full-loop helper

sync_queries(...) returns a typed QuerySyncSummary dataclass.

Basic usage:

queries = client.list_queries(include_all=True)
query_id = str(queries[0]["id"])

definition = client.get_query(query_id)
snapshot = client.execute_query(query_id)
incremental = client.get_query_data(
    query_id,
    if_newer_than_watermark=snapshot.get("source_watermark"),
)

Updated-since two-gate sync (recommended)

Use a two-gate strategy to avoid heavy query calls when no fresh data exists.

1. Definition gate (cheap): call list_queries(updated_since=...).
2. Data gate (targeted): for changed query IDs, call get_query_data(if_newer_than_watermark=...).

Query Sync Steps

1. Read persisted state:
   • last_sync_at (team-level ISO timestamp)
   • query_watermarks (map of query_id -> source_watermark)
2. Call list_queries(include_all=True, updated_since=last_sync_at).
3. If last_sync_at is empty (first sync), call without updated_since to fetch all query definitions once.
4. If response is empty, stop (no heavy calls).
5. For each returned query:
   • read previous watermark for that query
   • call get_query_data(query_id, if_newer_than_watermark=prev)
   • if payload has unchanged=True, skip downstream work
   • else process rows and persist returned source_watermark
6. Advance last_sync_at to the max updated_at returned in step 2.

End-to-end example

from datetime import datetime, timezone

state = client.get_core_state()
last_sync_at = state.get("last_sync_at")
query_watermarks = state.get("query_watermarks") or {}

changed_queries = client.list_queries(
    include_all=True,
    updated_since=last_sync_at,
)

if not changed_queries:
    print("No query definition updates since last sync.")
else:
    latest_updated_at = last_sync_at
    for query in changed_queries:
        query_id = str(query["id"])
        previous_watermark = query_watermarks.get(query_id)

        data = client.get_query_data(
            query_id,
            if_newer_than_watermark=previous_watermark,
        )

        if data.get("unchanged") is True:
            continue

        rows = data.get("rows") or []
        # TODO: upsert rows into external system

        watermark = data.get("source_watermark")
        if isinstance(watermark, str) and watermark:
            query_watermarks[query_id] = watermark

        updated_at = query.get("updated_at")
        if isinstance(updated_at, str) and (
            latest_updated_at is None or updated_at > latest_updated_at
        ):
            latest_updated_at = updated_at

    next_state = {
        "query_watermarks": query_watermarks,
        "last_sync_at": latest_updated_at
        or datetime.now(timezone.utc).isoformat(),
    }
    client.patch_core_state(next_state)

Query Convenience Method

def handle_query_result(query: dict, result: dict) -> None:
    if result.get("unchanged") is True:
        return
    rows = result.get("rows") or []
    # TODO: sync rows to external system

summary = client.sync_queries(
    persist_state=True,
    on_query_result=handle_query_result,
)
print(summary)

Reliable large payload sync (recommended)

For large writes (for example 70,000+ rows), use chunked sync with retry/backoff:

from teicor_sdk import TeicorClient

client = TeicorClient(
    base_url="https://api.teicor.com",
    team_slug="client-a",
    app_slug="shopify",
    service_token="tkmkt_...",
)

def on_chunk(result, response):
    print(
        f"chunk={result.chunk_index + 1} "
        f"size={result.chunk_size} attempts={result.attempts}"
    )

summary = client.sync_records_bulk(
    table_id="<table-id>",
    records=records,                 # list[dict]
    on_conflict="overwrite",        # idempotent upsert behavior
    chunk_size=500,                  # runtime default max
    max_retries=5,
    retry_base_seconds=0.5,
    retry_max_seconds=8.0,
    retry_jitter_ratio=0.2,
    on_chunk_complete=on_chunk,
)

print(summary)

Resume from core-state checkpoint after interruption:

checkpoint_state = client.get_bulk_sync_checkpoint()

def save_checkpoint(checkpoint):
    client.save_bulk_sync_checkpoint(checkpoint=checkpoint)

summary = client.sync_records_bulk(
    table_id="<table-id>",
    records=records,
    chunk_size=500,
    start_chunk_index=client.parse_bulk_sync_checkpoint(checkpoint_state),
    on_checkpoint=save_checkpoint,
)

if summary.records_processed == summary.total_records:
    client.clear_bulk_sync_checkpoint()

Single-call convenience wrapper (load/save/clear checkpoint automatically):

summary = client.sync_records_bulk_with_core_checkpoint(
    table_id="<table-id>",
    records=records,
    chunk_size=500,
    on_conflict="overwrite",
)

Notes:

• Keep deterministic unique/primary-key values so retries are safe.
• Keep chunk size at or below runtime max bulk setting (default 500).
• Use summary/chunk callbacks to checkpoint progress in your app.

Table two-gate sync (recommended)

Table sync follows the same two-gate pattern:

1. Table definition gate via list_tables(updated_since=...)
2. Table data gate via get_table_sync_status(if_newer_than_watermark=...)

SDK helpers:

• list_tables(updated_since=None)
• get_table_sync_status(table_id, if_newer_than_watermark=None)
• list_records(table_id=..., ...) for the actual row fetch after gates pass
• sync_tables(...) convenience full-loop helper

sync_tables(...) returns a typed TableSyncSummary dataclass.

Table Sync Steps

1. Read state:
   • tables_last_sync_at (team-level timestamp)
   • table_watermarks (table_id -> source_watermark)
2. Call list_tables(updated_since=tables_last_sync_at).
3. If tables_last_sync_at is empty (first sync), call without updated_since to discover all tables once.
4. If empty, stop.
5. For each changed table:
   • call get_table_sync_status(table_id, if_newer_than_watermark=...)
   • if has_changes is false, skip row fetch
   • if has_changes is true, call list_records(...) and process rows
   • persist returned source_watermark for that table
6. Advance tables_last_sync_at to max table updated_at from step 2.

Example

state = client.get_core_state()
tables_last_sync_at = state.get("tables_last_sync_at")
table_watermarks = state.get("table_watermarks") or {}

changed_tables = client.list_tables(updated_since=tables_last_sync_at)

if changed_tables:
    latest_table_updated_at = tables_last_sync_at

    for table in changed_tables:
        table_id = str(table["id"])
        prev_watermark = table_watermarks.get(table_id)

        status = client.get_table_sync_status(
            table_id,
            if_newer_than_watermark=prev_watermark,
        )
        if not status.get("has_changes"):
            continue

        # Fetch rows only when gate 2 says table data changed.
        records = client.list_records(table_id=table_id, limit=200, offset=0)
        # TODO: sync records into external system

        next_watermark = status.get("source_watermark")
        if isinstance(next_watermark, str) and next_watermark:
            table_watermarks[table_id] = next_watermark

        updated_at = table.get("updated_at")
        if isinstance(updated_at, str) and (
            latest_table_updated_at is None
            or updated_at > latest_table_updated_at
        ):
            latest_table_updated_at = updated_at

    client.patch_core_state(
        {
            "tables_last_sync_at": latest_table_updated_at,
            "table_watermarks": table_watermarks,
        }
    )

Table Convenience Method

def handle_table_records(table: dict, records: list[dict]) -> None:
    # TODO: upsert records into external system
    pass

summary = client.sync_tables(
    records_page_size=200,
    persist_state=True,
    on_table_records=handle_table_records,
)
print(summary)

Package maintainers

Build artifacts:

python -m pip install --upgrade build
python -m build

Publish to PyPI:

python -m pip install --upgrade twine
python -m twine upload dist/*

Auth headers used

• Authorization: Bearer <service_token>
• X-Teicor-App-Slug: <app_slug>

Notes

• core/state is app-scoped per team installation.
• Runtime calls default to Teicor's proxy; eligible installs can use short-lived direct runtime authority tokens.
• Source package is in sdk/python/src/teicor_sdk.