Changelog

Adds 9 customer-facing endpoints that the Atoms backend already serves but were not exposed in the OpenAPI spec, plus an x-fern-sdk-method-name cleanup pass on the phone-numbers and knowledge-base namespaces.

New endpoints

• POST /agent/with-ai — create an agent from a natural-language brief.
• GET /agent/{id}/call-logs — fetch conversation logs for an agent.
• GET /agent/{id}/widget-config — read the chat-widget configuration for an agent.
• PATCH /agent/{id}/widget-config — update the chat-widget configuration for an agent.
• GET /agent/prompt-config — fetch prompt-config metadata (question definitions + labels) used by the agent builder.
• GET /dnc — list DNC entries for the organization.
• GET /account/get-account-details — get the authenticated user + organizations they belong to.
• PATCH /account/update-org-name — update the organization’s display name.
• GET /user/subscription — get the caller’s subscription, plan limits, and feature flags.
• GET /campaign/{id}/logs/export — export a campaign’s call logs as a JSON file.

SDK method renames

x-fern-sdk-{group,method}-name annotations added across the phone_numbers and knowledge_base namespaces so the generated client uses canonical short names instead of summary-derived verbose ones.

Behavior change for SDK consumers

After the next regen, each of the 9 endpoints is callable via the auto-generated client (under client.atoms.agents.*, client.atoms.dnc.*, etc.). No changes to existing endpoint paths or request/response shapes.

New agent endpoints (7). New endpoints for creating, configuring, and operating agents:

Agent duplicate endpoint restructured. POST /agent/{id}/duplicate is now clearly documented at its own path. Logic is unchanged — pass targetOrganizationId (24-char hex ObjectId) to copy an agent into another org. Returns 403 if the caller is not a member of the target org, 404 if the agent or org is not found.

Draft diff response typed. The data field on GET /agent/{id}/drafts/{draftId}/diff is now typed with explicit fields: sections (keyed by section name, each with before, after, and changed) and hasChanges: boolean.

New analytics endpoints (18). A full analytics surface is now documented under the /analytics prefix:

New call actions endpoints (full CRUD). Call actions are automated behaviors fired at call lifecycle events. The action body includes a config.conditions[] array using the new CallActionCondition schema (field, operator, value):

New concurrency endpoints.

New integrations endpoints. Both endpoints return a direct JSON body rather than the standard { success, data } envelope:

New endpoint: disposition metric templates.

→ Atoms API reference

Spec changes that prepare the next Python SDK regen so the auto-generated client (a) preserves the 4.4.7 sub-package import surface that customers still rely on and (b) picks up several missing or mistyped fields from the live API.

Generator pin

fern-python-sdk is pinned to 4.61.3 (the version 4.4.x was generated with). 5.12.12 combined with exclude_types_from_init_exports: true strips type exports from every package __init__, not just the top-level one — Fern’s own auto-generated wire tests do from smallestai.atoms.<ns> import <Type> and ImportError out. The pin is the proper fix until Fern ships a 5.x build that scopes the exclusion to top-level only.

Spec field/type fixes

• listAgentTemplates — _id vs id: the _id (ObjectId) and the human id slug are distinct. Fern was aliasing _id → id and the two fields collided. Added x-fern-property-name: object_id on _id so the generated SDK exposes .object_id and .id as separate attributes. Same fix on singlePromptConfig._id.
• Template tools[].type: added the missing enum (end_call | transfer_call | api_call | extract_dynamic_variables | knowledge_base_search).
• live_transcripts.metrics: was typed object with additionalProperties: true; server actually sends an array of {processor, model, value} objects. The SDK was silently dropping 122 SSE metrics events in a 40-second call before this. Now typed as an array with the correct item shape.
• calls.list / calls.search — turnLatencyMetrics: brought in line with calls.get by adding minLatency, maxLatency, transitions[], and processedAt.

Method-name cleanup

20 operations had verbose summary-derived method names; added explicit x-fern-sdk-group-name + x-fern-sdk-method-name annotations so the SDK surface stays readable. The worst offender — edit_draft_config_prompt_tools_post_call_metrics_voice_etc — is now agent_versioning_drafts.update_draft_config.

Renames cover: campaigns (list/create/get/delete/start_or_resume/pause), compliance (submit/resubmit), webhooks (create/delete), audience (list/get), agent_versioning_drafts (create_draft/rename_draft/discard_draft/publish_draft/update_draft_config), and agent_versioning_versions (compare_version_metrics/update_version_metadata/activate_version).

Synthesizer model enum

agent_dto.synthesizer.voiceConfig.model (and the template variant) now lists waves_lightning_v3_1 first and uses it as the default (was waves_lightning_large). The deprecated tokens (waves, waves_lightning_large, waves_lightning_large_voice_clone, waves_lightning_v2, waves_lightning_v3, waves_lightning_v2_http) remain in the enum for backwards compatibility — existing agents using them keep working — but the description flags them and notes removal once the new lineup tokens (Lightning v3.1 Pro, Pulse, Pulse Pro, Hydra) are confirmed and added.

Behavior change for SDK consumers

After the next regen lands:

• agent.object_id returns the ObjectId for an agent template (previously colliding with the slug).
• live_transcripts event handlers that read metrics need to iterate it as a list of objects, not look up keys on a dict.
• The 4.4.x verbose method names on the operations listed above are gone — only the short names work.

No customer-facing endpoint paths or request shapes change.

smallestai==5.1.0 adds 10 endpoints that were already serving traffic but missing from the SDK surface, cleans up auto-generated method names across the namespaces that still had verbose summary-derived names, and surfaces 19 customer-facing fields on the agent response that the auto-generated client previously couldn’t type. Constructor signature is unchanged from 5.0.0 — SmallestAI(api_key="...") keeps working.

Update your install

$
pip install --upgrade smallestai
$
# pinned: pip install smallestai==5.1.0

New endpoints

Method renames

Verbose summary-derived method names → canonical short names. Applied across 20 operations:

The 5.0.x verbose names are removed in 5.1.0 with no deprecation alias.

Newly-typed agent response fields

GET /agent/{id} was already returning these 19 fields at runtime, but the spec didn’t include them — so the 5.0.x client couldn’t surface them as typed attributes. Each field is now a typed attribute on the agent response:

allow_inbound_call, phone_number, visible_to_everyone, speech_formatting, mute_user_until_first_bot_response, interruption_backoff_timer, enable_style_guide, call_disposition_config, voice_mail_detection_config, smart_turn_config, voice_detection_config, denoising_config, redaction_config, pronunciation_dicts, llm_idle_timeout_config, session_timeout_config, timezone, post_call_analytics_config, widget_config.

Each shape is documented in the agent response schema.

Other spec corrections

• agent_templates — _id (stable identifier) and id (human-readable slug) are now distinct on the SDK side. _id surfaces as template.object_id; id stays as template.id. Previous releases aliased one over the other.
• live_transcripts.metrics — typed as a list of {processor, model, value} items (was untyped). The 5.0.x client silently dropped these events on deserialization.
• calls.list / calls.search — turn_latency_metrics now includes min_latency, max_latency, transitions, processed_at (parity with calls.get).

Migration

The rename pattern is mechanical. A single repo-wide sed:

$
find . -name '*.py' -exec sed -i.bak \
>
  -e 's/get_all_campaigns/list_campaigns/g' \
>
  -e 's/start_or_resume_campaign/start_or_resume/g' \
>
  -e 's/pause_campaign/pause/g' \
>
  -e 's/submit_a_compliance_application/submit_compliance/g' \
>
  -e 's/resubmit_a_rejected_compliance_application/resubmit_compliance/g' \
>
  -e 's/create_a_webhook/create_webhook/g' \
>
  -e 's/delete_a_webhook/delete_webhook/g' \
>
  -e 's/get_all_audiences/list_audiences/g' \
>
  -e 's/get_audience_by_id/get_audience/g' \
>
  -e 's/edit_draft_config_prompt_tools_post_call_metrics_voice_etc/update_draft_config/g' \
>
  -e 's/compare_metrics_between_two_versions/compare_version_metrics/g' \
>
  -e 's/update_version_metadata_label_description_pin_only/update_version_metadata/g' \
>
  -e 's/activate_a_version/activate_version/g' \
>
  {} \;

Full release notes and the SDK-side migration write-up: MIGRATION.md on smallest-python-sdk.

PR #262 pinned the Python SDK generator from 5.12.12 down to 4.61.3 to address sub-package __init__ type-export stripping under exclude_types_from_init_exports: true. An empirical regen test against 4.61.3 showed that change was the wrong fix:

• 4.61.3’s core/ is missing helpers (encode_path_param, _make_default_async_client) that our preserved / .fernignore’d files depend on — so the regen produces mypy failures it didn’t have on 5.12.12.
• 4.61.3 also does not restore the sub-package type exports the downgrade was supposed to fix — the empty __init__ symptom reproduces on it.

The actual fix for preserving the sub-package import surface lives in the SDK repo itself (.fernignore the 3 sub-package __init__s with their hand-written type exports — waves is already done; webhooks and prompt_scoring will follow). That keeps the generator on a coherent 5.12.12 core/ with no version skew.

All other PR #262 / #263 spec changes (the 20 method-name renames, synthesizer enum cleanup, _id / object_id rename, live_transcripts.metrics array fix, calls turn-latency fields, 19 new AgentDTO fields) are generator-independent and stay as-is.