Seclai Python SDK

The official Python SDK for the Seclai API. Provides typed wrappers for the Seclai API, file uploads, SSE streaming, polling helpers, and full async support.

Requires Python 3.11+.

Install

pip install seclai

Exports

All public symbols are available from the top-level seclai package:

from seclai import (
    Seclai,                     # Synchronous client
    AsyncSeclai,                # Asynchronous client
    SeclaiError,                # Base exception
    SeclaiConfigurationError,   # Missing API key / invalid config
    SeclaiAPIStatusError,       # Non-2xx HTTP response
    SeclaiAPIValidationError,   # HTTP 422 validation error
    SeclaiStreamingError,       # SSE stream error event
    AgentRunStreamRequest,      # TypedDict for streaming run requests
    JSONValue,                  # Recursive JSON type alias
)

Quick start

from seclai import Seclai

client = Seclai(api_key="...")

# List agents
agents = client.list_agents()
print(agents)

# Run an agent and stream the result
from seclai import AgentRunStreamRequest

run = client.run_streaming_agent_and_wait(
    "agent_id",
    body=AgentRunStreamRequest(input="Summarize the latest uploads", metadata={}),
    timeout=60.0,
)
print("run:", run.run_id, "status:", run.status)

Async client

import asyncio
from seclai import AsyncSeclai

async def main():
    async with AsyncSeclai(api_key="...") as client:
        agents = await client.list_agents()
        print(agents)

asyncio.run(main())

Configuration

Set SECLAI_API_URL to point at a different API host (e.g., staging):

export SECLAI_API_URL="https://staging-api.seclai.com"

Authentication

Credentials are resolved via a chain (first match wins):

1. Explicit api_key option
2. Explicit access_token option (string or callable)
3. SECLAI_API_KEY environment variable
4. SSO — cached tokens from ~/.seclai/sso/cache/ (always available as fallback)

# API key
client = Seclai(api_key="sk-...")

# Static bearer token
client = Seclai(access_token="eyJhbGciOi...")

# Dynamic bearer token provider (sync callable, called per request)
client = Seclai(access_token=lambda: get_token_from_vault())

# Async provider — use AsyncSeclai for async callables
client = AsyncSeclai(access_token=get_token_async)

# SSO profile (uses cached tokens, auto-refreshes)
client = Seclai(profile="my-profile")

# Environment variable (no options needed)
# export SECLAI_API_KEY="sk-..."
client = Seclai()

SSO authentication

SSO is the default fallback when no explicit credentials are provided. The SDK includes built-in production SSO defaults, so no configuration is needed:

npx @seclai/cli auth login    # authenticate via browser — works immediately

To customize SSO settings (e.g. for a staging environment), use seclai configure sso or set environment variables:

API documentation

Online API documentation (latest):

https://seclai.github.io/seclai-python/1.7.0/

API versioning

The API dates its backward-incompatible changes. Nothing changes for you until you opt in, either per client or by pinning the account:

from seclai import ApiVersion

client = Seclai(api_key="...", api_version=ApiVersion.V2026_07_27)  # Seclai-Version header

state = client.get_api_version()          # what this request resolved to
client.update_api_version(ApiVersion.V2026_07_27)  # pin the whole account

Leave api_version unset and the header is omitted, so the account's pinned baseline applies and responses keep their current shapes. Upgrading this package alone never changes the wire contract.

Known versions are on the ApiVersion string enum, alongside DEFAULT_API_VERSION and LATEST_API_VERSION. A version this release was not built against raises SeclaiConfigurationError: a newer version can reshape responses, and this client would decode them incorrectly rather than reject them. Upgrade the package to adopt a new version, or pass allow_unknown_api_version=True if you have to move first and accept that risk.

The guard covers the header however it is supplied — api_version, default_headers, or a per-request headers argument — and nothing else. An account pinned server-side can still be newer than this release — get_api_version() reports the effective_version the request resolved to, and comparing it against LATEST_API_VERSION is how you detect the gap.

What 2026-07-27 changes. Undeclared query parameters become a 422 instead of being ignored, and list endpoints move to the canonical {"data": [...], "pagination": {...}} envelope. The affected methods read both shapes, so they keep working either way — but the metadata moves:

unwrap_items() reads either shape, so a call site does not have to branch on the version:

from seclai import unwrap_items

items = unwrap_items(client.list_alert_configs(), "configs")
items = unwrap_items(client.list_model_alerts(), "alerts")

Prefer pagination over the flat keys. The legacy keys will be deprecated and then removed once the canonical envelope is the default.

The two evaluation endpoints are unpaginated on the legacy shape — they ignore page/limit and return everything — so a paginate-until-empty loop over them only terminates once you have opted in. get_generation_tiers() takes no paging arguments at all and always returns the full set. The remaining three paginate on either shape.

Later versions. Each is cumulative, and none changes a response shape this client decodes:

The cloud-drive listings and list_embedding_models() / list_reranker_models() follow the 2026-07-27 envelope rule as well. The cloud-drive methods return the items on either shape; read the two model listings with unwrap_items(result, "models").

Resources

Identity

me = client.get_me()
print(me["account_id"])
for org in me["organizations"]:
    print(org["name"], org["account_id"])

# Act as an organization: Seclai(account_id=org["account_id"])

Agents

# CRUD
agents = client.list_agents(page=1, limit=20)
agent = client.create_agent({"name": "My Agent", "description": "..."})
fetched = client.get_agent("agent_id")

# Pause / resume — a disabled agent stops firing from every trigger path
callers = client.get_agent_callers("agent_id")  # live agents calling this one
client.disable_agent("agent_id")  # 409 if any caller above is still live
client.enable_agent("agent_id")
updated = client.update_agent("agent_id", {"name": "Renamed"})
client.delete_agent("agent_id")

# Definition (step workflow)
definition = client.get_agent_definition("agent_id")
client.update_agent_definition("agent_id", {
    "change_id": definition["change_id"],
    "steps": [{"type": "llm", "config": {}}],
})

# Export / import an agent
exported = client.export_agent("agent_id")

# Validate the payload first to surface unresolved entity refs in this account
preview = client.preview_import_agent({"agent_definition": exported})
entity_remap = {
    ref["ref_id"]: ""  # pick a target uuid from ref["alternatives"]
    for ref in preview.get("unresolved_refs", [])
}

# Commit — `entity_remap` substitutes workflow refs before save
imported = client.create_agent({
    "name": "Imported",
    "agent_definition": exported,
    "entity_remap": entity_remap,
})
# `imported["import_warnings"]` lists any items that couldn't be applied.

Agent runs

from seclai._generated.models.agent_run_request import AgentRunRequest

# Start a run
run = client.run_agent("agent_id", AgentRunRequest(input_="Hello"))

# List & search runs
runs = client.list_agent_runs("agent_id")
search = client.search_agent_runs({"query": "test"})

# Fetch run details (optionally with step outputs)
detail = client.get_agent_run("run_id", include_step_outputs=True)

# Cancel an in-flight or queued run. `delete_agent_run()` is the same
# operation on the same endpoint, returning a typed model instead of a dict.
client.cancel_agent_run("run_id")

Streaming

The SDK provides two streaming patterns over the SSE /runs/stream endpoint.

Block until done — returns the final done payload or raises on timeout:

from seclai import AgentRunStreamRequest

run = client.run_streaming_agent_and_wait(
    "agent_id",
    body=AgentRunStreamRequest(input="Hello from streaming", metadata={}),
    timeout=60.0,
)

Generator-based — yields every SSE event as (event_type, data) tuples:

for event_type, data in client.run_streaming_agent(
    "agent_id",
    body=AgentRunStreamRequest(input="Hello", metadata={}),
):
    print(event_type, data)

Async:

async for event_type, data in client.run_streaming_agent(
    "agent_id",
    body=AgentRunStreamRequest(input="Hello", metadata={}),
):
    print(event_type, data)

Polling

For environments where SSE is not practical, poll for a completed run:

from seclai._generated.models.agent_run_request import AgentRunRequest

result = client.run_agent_and_poll(
    "agent_id",
    AgentRunRequest(input_="Hello"),
    poll_interval=2.0,
)

Agent input uploads

# Discover which files (if any) the agent expects before staging uploads
refs = client.get_agent_attachment_references("agent_id")
# refs["requires_uploads"] -> bool; refs["agent"] lists the exact_names /
# indexes_max / patterns a run-time upload batch must satisfy.

upload = client.upload_agent_input("agent_id", file=b"data", file_name="input.pdf")
status = client.get_agent_input_upload_status("agent_id", upload["upload_id"])

Agent run attachments

# Download a file emitted by a step in an agent run. attachment_id is the
# URL-safe-base64 storage_key surfaced in run output manifests / webhooks.
response = client.download_agent_run_attachment("run_id", "attachment_id")  # raw httpx.Response
with response:
    for chunk in response.iter_bytes():
        ...  # write to disk

Agent AI assistant

steps = client.generate_agent_steps("agent_id", {"user_input": "Build a RAG pipeline"})
config = client.generate_step_config("agent_id", {"step_type": "llm", "user_input": "..."})

# Conversation history — step_type is required by the API
history = client.get_agent_ai_conversation_history("agent_id", step_type="llm")
client.mark_agent_ai_suggestion("agent_id", "conversation_id", {"accepted": True})

Agent evaluations

# CRUD
criteria_list = client.list_evaluation_criteria("agent_id", page=1, limit=50)
# page/limit only take effect with api_version="2026-07-27" or later; the legacy
# response is unpaginated. list_evaluation_criteria_page() returns the same items
# plus a "pagination" key when opted in.
criteria = client.create_evaluation_criteria("agent_id", {"name": "accuracy"})
detail = client.get_evaluation_criteria("criteria_id")
client.update_evaluation_criteria("criteria_id", {"name": "updated"})
client.delete_evaluation_criteria("criteria_id")

# Test a draft
client.test_draft_evaluation("agent_id", {"criteria": {}, "run_id": "run_id"})

# Results & summaries
results = client.list_evaluation_results("criteria_id")
summary = client.get_evaluation_criteria_summary("criteria_id")
client.create_evaluation_result("criteria_id", {"run_id": "run_id", "score": 0.9})

# Results by run
run_results = client.list_run_evaluation_results("agent_id", "run_id")
non_manual = client.get_non_manual_evaluation_summary("agent_id")
compatible = client.list_compatible_runs("criteria_id")

Knowledge bases

kbs = client.list_knowledge_bases()
kb = client.create_knowledge_base({"name": "My KB"})
fetched = client.get_knowledge_base("kb_id")
client.update_knowledge_base("kb_id", {"name": "Renamed"})
client.delete_knowledge_base("kb_id")

Memory banks

banks = client.list_memory_banks()
bank = client.create_memory_bank({"name": "Chat Memory", "type": "conversation"})
fetched = client.get_memory_bank("mb_id")
client.update_memory_bank("mb_id", {"name": "Updated"})
client.delete_memory_bank("mb_id")

# Stats & compaction
stats = client.get_memory_bank_stats("mb_id")
client.compact_memory_bank("mb_id")

# Test compaction
test = client.test_memory_bank_compaction("mb_id", {"entries": []})
standalone = client.test_compaction_prompt_standalone({"prompt": "test"})

# Templates & agents
templates = client.list_memory_bank_templates()
agents = client.get_agents_using_memory_bank("mb_id")

# AI assistant
suggestion = client.generate_memory_bank_config({"user_input": "Create a bank"})
last_conv = client.get_memory_bank_ai_last_conversation()
client.accept_memory_bank_ai_suggestion("conversation_id", {"accepted": True})

# Source management
client.delete_memory_bank_source("mb_id")

Sources

sources = client.list_sources(page=1, limit=20)
source = client.create_source({"name": "My Source"})
fetched = client.get_source("source_id")
client.update_source("source_id", {"name": "Updated"})
client.delete_source("source_id")

Indexing status of a source's content, keyed by the content_version_id the upload methods return:

failed = client.list_source_contents("source_id", status="failed")
batch = client.list_source_contents(
    "source_id", content_version_ids=["cv_1", "cv_2"]
)
one = client.get_source_content_status("source_id", "cv_1")

Cloud drives

providers = client.list_cloud_drive_providers()
drives = client.list_cloud_drives()
drive = client.get_cloud_drive("connection_id")
client.update_cloud_drive("connection_id", {"name": "Contracts"})

# Which agents depend on it, and which files it skipped and why
agents = client.get_agents_using_cloud_drive("connection_id")
skipped = client.list_cloud_drive_rejections("connection_id", limit=20)

client.disconnect_cloud_drive("connection_id")  # keeps the connection
client.delete_cloud_drive("connection_id")

File uploads

Upload a file to a source (max 200 MiB):

upload = client.upload_file_to_source(
    "source_connection_id",
    file="./document.pdf",
    title="Q4 Report",
    metadata={"department": "finance"},
)

Upload inline text:

upload = client.upload_inline_text_to_source("source_connection_id", {
    "title": "Greeting",
    "content": "Hello, world!",
})

Replace a content version with a new file:

upload = client.upload_file_to_content(
    "source_connection_content_version",
    file="./updated.pdf",
    metadata={"revision": 2},
)

Replace a content version with inline text:

client.replace_content_with_inline_text("source_connection_content_version", {
    "title": "Updated",
    "content": "New content text",
})

Source exports

exports = client.list_source_exports("source_id")
export = client.create_source_export("source_id", {"format": "json"})
status = client.get_source_export("source_id", "export_id")
estimate = client.estimate_source_export("source_id", {"format": "json"})
response = client.download_source_export("source_id", "export_id")  # raw httpx.Response
client.delete_source_export("source_id", "export_id")
client.cancel_source_export("source_id", "export_id")

Source embedding migrations

migration = client.get_source_embedding_migration("source_id")
client.start_source_embedding_migration("source_id", {"target_model": "v2"})
client.cancel_source_embedding_migration("source_id")

Content

detail = client.get_content_detail("source_connection_content_version")
embeddings = client.list_content_embeddings("source_connection_content_version")
client.delete_content("source_connection_content_version")

Solutions

solutions = client.list_solutions()
sol = client.create_solution({"name": "My Solution"})
fetched = client.get_solution("solution_id")
client.update_solution("solution_id", {"name": "Renamed"})
client.delete_solution("solution_id")

# Link / unlink resources
client.link_agents_to_solution("solution_id", {"agent_ids": ["a1"]})
client.unlink_agents_from_solution("solution_id", {"agent_ids": ["a1"]})
client.link_knowledge_bases_to_solution("solution_id", {"kb_ids": ["kb1"]})
client.unlink_knowledge_bases_from_solution("solution_id", {"kb_ids": ["kb1"]})
client.link_source_connections_to_solution("solution_id", {"sc_ids": ["sc1"]})
client.unlink_source_connections_from_solution("solution_id", {"sc_ids": ["sc1"]})

# AI assistant
plan = client.generate_solution_ai_plan("solution_id", {"user_input": "Build it"})
client.accept_solution_ai_plan("solution_id", "conversation_id", {})
client.decline_solution_ai_plan("solution_id", "conversation_id")

# AI-generated resources
client.generate_solution_ai_knowledge_base("solution_id", {"user_input": "..."})
client.generate_solution_ai_source("solution_id", {"user_input": "..."})

# Conversations
convs = client.list_solution_conversations("solution_id")
client.add_solution_conversation_turn("solution_id", {"user_input": "..."})
client.mark_solution_conversation_turn("solution_id", "conversation_id", {"accepted": True})

Governance AI

plan = client.generate_governance_ai_plan({"user_input": "Create a content policy"})
convs = client.list_governance_ai_conversations()
client.accept_governance_ai_plan("conversation_id")
client.decline_governance_ai_plan("conversation_id")

Alerts

alerts = client.list_alerts(status="active")
alert = client.get_alert("alert_id")
client.change_alert_status("alert_id", {"status": "resolved"})
client.add_alert_comment("alert_id", {"text": "Investigating"})

# Subscriptions
client.subscribe_to_alert("alert_id")
client.unsubscribe_from_alert("alert_id")

# Alert configs
configs = client.list_alert_configs()
client.create_alert_config({"name": "Config"})
config = client.get_alert_config("config_id")
client.update_alert_config("config_id", {"name": "Updated"})
client.delete_alert_config("config_id")

# Organization preferences
prefs = client.list_organization_alert_preferences()
client.update_organization_alert_preference("org_id", "anomaly", {"enabled": True})

Agent email triggers

# Configure an EMAIL_RECEIVED trigger; omitted fields are left unchanged
config = client.set_email_trigger_config(
    "agent_id",
    "trigger_id",
    {
        "alias": "support",
        "allowed_senders": ["example.com", "ops@partner.com"],
        "ignore_auto_generated": True,   # drop auto-replies to prevent loops
        "require_sender_auth": True,     # require SPF or DMARC
        "queue_on_quota": False,         # park over-rate mail instead of failing
    },
)
print(config["email_addresses"])

Agent email governance

# Recipients who opted out of this account's agent emails
opt_outs = client.list_agent_email_optouts(agent_id="agent_id", limit=50)
client.remove_agent_email_optout("optout_id")  # opt them back in

# Blocked inbound senders (owner/admin only)
blocked = client.list_blocked_email_senders(limit=50)
client.block_email_sender({"sender_email": "spam.example.com", "match_type": "domain"})
client.unblock_email_sender("blocked_id")

# Auto-block on a governance BLOCK: "disabled" | "input" | "input_and_output"
client.set_auto_block_mode({"mode": "input_and_output"})

# Inbound mail discarded before running an agent
rejections = client.list_inbound_email_rejections(agent_id="agent_id")

# Account-wide overload circuit breaker
status = client.get_inbound_email_status()  # {"paused": ..., "queued_backlog": ...}
client.cancel_queued_email_runs()  # fail all QUEUED (over-quota parked) runs
client.resume_inbound_email()      # one-shot; re-arms if still overloaded

Email domains

Send and receive agent email on your own domain instead of the shared agent.seclai.com. Requires a user-bound credential; mutations require an account owner/admin.

listing = client.list_email_domains()

vanity = client.add_email_domain({"kind": "vanity", "value": "acme"})
custom = client.add_email_domain(
    {"kind": "custom", "value": "agent.mycompany.com", "delegated": True}
)

# Publish custom["dns_records"], then check without waiting for the sweep
client.verify_email_domain(custom["id"])

client.set_primary_email_domain(custom["id"])
client.use_shared_email_domain()  # revert; domains stay configured & verified

client.send_email_domain_test_email(custom["id"])  # always to the account owner
dmarc = client.get_dmarc_summary(custom["id"], days=30, top_sources=10)

removed = client.remove_email_domain(custom["id"])
print(removed.get("cleanup_note"))  # set when the domain was Seclai-managed

Models

# Media-generation quality tiers (fast/balanced/thorough) and what each resolves to
tiers = client.get_generation_tiers()

# Embedding and reranker models, with their pricing
embedders = unwrap_items(client.list_embedding_models(), "models")
rerankers = unwrap_items(client.list_reranker_models(), "models")

alerts = client.list_model_alerts()
client.mark_model_alert_read("alert_id")
client.mark_all_model_alerts_read()
unread = client.get_unread_model_alert_count()
recs = client.get_model_recommendations("model_id")

# Model playground experiments
experiment = client.create_experiment({"model_ids": ["model_id"], "prompt": "..."})
experiments = client.list_experiments()
detail = client.get_experiment("experiment_id")
client.cancel_experiment("experiment_id")
client.delete_experiment("experiment_id")  # soft-delete, preserves audit history

Search

results = client.search(query="quarterly report")
filtered = client.search(query="my agent", entity_type="agent", limit=5)

Documentation search

Results are global (not account-scoped); each carries a doc_slug plus an optional anchor for building a https://seclai.com/docs/<doc_slug>[#<anchor>] link.

hits = client.search_docs("email triggers")                       # fast keyword match
deep = client.search_docs("how do I stop auto-reply loops",
                          mode="semantic", limit=5)               # adds a highlight

Top-level AI assistant

# Generate plans for different resource types
kb_plan = client.ai_assistant_knowledge_base({"user_input": "Create a product FAQ KB"})
source_plan = client.ai_assistant_source({"user_input": "Set up a docs source"})
solution_plan = client.ai_assistant_solution({"user_input": "Build a support bot"})
mb_plan = client.ai_assistant_memory_bank({"user_input": "Create a chat memory bank"})

# Accept or decline
client.accept_ai_assistant_plan("conversation_id", {"accepted": True})
client.decline_ai_assistant_plan("conversation_id")

# Memory bank conversation history
history = client.get_ai_assistant_memory_bank_history()
client.accept_ai_memory_bank_suggestion("conversation_id", {"accepted": True})

# Feedback
client.submit_ai_feedback({"rating": 5, "comment": "Helpful!"})

Pagination

All list methods accept page and limit parameters. For auto-pagination across all pages, use the paginate helper:

# Sync — yields items one by one (generator)
for agent in client.paginate("GET", "/agents"):
    print(agent["name"])

# With a per-resource items key; `data` is read first, so this works on
# either response shape
for config in client.paginate("GET", "/alerts/configs", items_key="configs"):
    print(config["id"])

# Endpoints that declare `offset` rather than `page`
for alert in client.paginate("GET", "/models/alerts", items_key="alerts",
                             param_style="offset"):
    print(alert["id"])

# Async — also an async generator
async for agent in client.paginate("GET", "/agents"):
    print(agent["name"])

Error handling

All SDK errors inherit from SeclaiError. Use specific exception types for targeted handling:

from seclai import (
    Seclai,
    SeclaiAPIStatusError,
    SeclaiAPIValidationError,
    SeclaiConfigurationError,
    SeclaiStreamingError,
)

client = Seclai(api_key="...")

try:
    from seclai._generated.models.agent_run_request import AgentRunRequest

    result = client.run_agent("agent_id", AgentRunRequest(input_="Hello"))
except SeclaiAPIValidationError as e:
    print("Validation error:", e.status_code, e.validation_error)
except SeclaiAPIStatusError as e:
    print("API error:", e.status_code, e.response_text)
except SeclaiStreamingError as e:
    print("Streaming error:", e.message, "run:", e.run_id)
except SeclaiConfigurationError as e:
    print("Config error:", e)

Low-level access

Use client.request() for direct API requests:

result = client.request("GET", "/custom/endpoint", params={"key": "value"})

Development

Testing

make test

To pass args through to pytest:

make test ARGS='-k auth'

Formatting

make format

Linting

make lint

OpenAPI spec & regenerating the client

Copy the OpenAPI JSON file into openapi/seclai.openapi.json, then run:

make generate

Generate docs

make docs

Reporting issues

If you hit a bug or have a feature request, please open an issue and include:

• what you were trying to do
• a minimal repro snippet (if possible)
• the exception / traceback
• your environment (Python version, OS)

License

MIT — see LICENSE for details.