v0.2.14-ce

Summary

This release packages the changes delivered since v0.2.13-ce into the v0.2.14-ce release line.

Release Notes

• extends runtime configuration and provider coordination across core application services
• refines interface flows for administrative, onboarding, and service-facing routes
• updates orchestration, Matrix, governance, and Temporal coordination paths
• aligns repository release metadata and version references for v0.2.14-ce

Included Range

• d3a4896 Refine runtime fallback and configuration paths
• fd382c4 Refine runtime and interface boundaries
• 92176ff Bump version to v0.2.14-ce

v0.2.13-ce

Highlights

Automata v0.2.13-ce strengthens governance enforcement and improves self-hosted Kubernetes deployment support.

• Agent responses are now guarded by a law certification step that checks the chosen response against the active constitution snapshot before it is delivered.
• When a response cannot be certified, the workflow blocks the original output, falls back to a lawful message, and keeps certification data attached to the run for auditing.
• On-prem K3s deployment support is improved with Kubernetes Helm charts and NodePort service updates to make self-hosted installs easier to package and expose.

Included changes

• Add law certification guard
• Add NodePort services for on-premises K8s deployment
• Add K8s Helm charts for k3s deployment

v0.2.12-ce

Release v0.2.12-ce

What's Changed

• Kubernetes Support: Added a production-ready, highly-configurable Helm chart for on-premise (e.g., k3s) deployments including Let's Encrypt support and Tailscale subchart integrations.
• NixOS Support: Replaced the previous Podman-based VM with a fully native flake.nix making it simple to deploy Automata on arbitrary NixOS configurations natively.
• Documentation (docs/OPERATIONS.md and docs/K8S_DEPLOYMENT.md) has been greatly expanded.

v0.2.11-ce

SentientWave Automata Community Edition v0.2.11-ce packages the latest provider work and README positioning updates since v0.2.10-ce.

Google Gemini provider support

• added native Google Gemini provider support in the Automata backend
• aligned Gemini request handling with the provider's system-instruction payload shape
• exposed Gemini as a configurable provider in the admin UI
• improved Gemini setup guidance in the admin console with provider-aware presets, endpoint hints, and credential guidance

Refreshed provider recommendations

• updated current recommended hosted-provider model selections across the stack
• refreshed admin UI provider guidance and placeholders so operators see current recommended model choices while configuring providers
• updated all-in-one quickstart defaults and operator docs for the refreshed hosted-provider recommendations
• kept local runtime defaults stable while still surfacing newer recommendations for local providers in the admin UI

Clearer README messaging

• rewrote the README opening to explain more clearly what Automata is and why it exists
• clarified Automata as a new operational layer for humans and AI agents in a world where software actors increasingly participate in work and decisions
• added stronger messaging around autonomy, self-evolution, and dynamic governance through laws and proposals
• improved onboarding clarity and top-level product framing

v0.2.10-ce

SentientWave Automata Community Edition v0.2.10-ce focuses on making the system more reliable in day to day operation and easier to understand from the first page a user sees.

The most important runtime work in this release hardens Temporal-backed execution. Conversation workflow persistence is now more robust when workflow activity payloads arrive in shapes that need normalization before storage. Permanent Matrix posting failures in conversation workflows are also handled more cleanly, so the system stops retrying forever when a room error is terminal instead of transient. Related non-retryable handling across agent, governance, and scheduled-task activity paths was tightened as part of the same reliability pass.

This release also improves Matrix reconciliation behavior for default room membership. When Synapse responds with M_LIMIT_EXCEEDED, Automata now respects the server-provided retry_after_ms window for default room invite and join attempts. That reduces repeated reconcile failures and log noise while still allowing membership reconciliation to recover cleanly when the homeserver is ready again.

On the admin side, the internal console language was cleaned up to feel simpler and more consistent. The UI now uses Providers instead of LLM Providers and Traces instead of LLM Traces, with matching updates across navigation, page titles, dashboard cards, flash messages, and controller-backed UI tests.

The README was also improved in two useful ways. It now includes a high-level explanation of how Automata uses Temporal as its durable workflow engine, and the top of the README now includes a website badge plus a current-version badge linked to the matching release.

Included changes in this release:

• stronger Temporal workflow retry handling and result persistence for conversation workflows
• cleaner handling of permanent Matrix room posting failures in Temporal-backed orchestration
• reduced Matrix reconcile noise by backing off default room membership retries on rate limiting
• simpler admin UI wording for provider and trace management
• clearer README explanation of Temporal in Automata
• website and version badges added to the README

v0.2.9-ce

SentientWave Automata Community Edition v0.2.9-ce delivers a major upgrade to how agentic work is executed and expands what agents can do when a goal needs deeper investigation.

The biggest change in this release is the workflow execution model. Agent runs, direct-message and mention-triggered runs, scheduled task execution, governance proposal lifecycles, and generic conversation workflows now run through the in-repo Elixir Temporal runtime instead of local fallback execution paths. This makes the system much closer to the durable execution model Automata is designed around, with Temporal owning the long-running workflow state and execution flow.

On top of that Temporal cutover, this release includes a fix for Temporal DM agent run execution so payload normalization behaves correctly under real workflow activity dispatch and retry conditions. That improves reliability for private conversations, which are one of the most important day-to-day interaction paths in Automata.

This release also introduces deep research workflow support for agents. When a goal is clearly more complex, needs fresh evidence, or the user explicitly asks for deep research, the agent workflow can branch into a bounded research path. That research path is implemented in Elixir and runs fully through Temporal activities. It assesses whether deeper research is warranted, executes web research rounds, reviews evidence between rounds, and synthesizes a final response based on the gathered material.

To support that research path, the Brave search tool was expanded so a single research round can execute multiple related queries and merge deduplicated evidence into a cleaner result set. This gives the workflow better raw material for review and synthesis without pushing the agent back into ad hoc, controller-side execution.

The all-in-one deployment template and runtime configuration now also include deep research limits for rounds, per-round queries, and results per query, so operators can tune how aggressively research workflows behave in local and self-hosted setups.

Included changes in this release:

• Temporal-only workflow execution cutover in the Elixir runtime
• stronger Temporal-backed handling for agent runs, scheduled tasks, governance workflows, and generic workflows
• DM execution reliability fix for Temporal activity payload handling
• new deep research workflow path for complex or explicitly research-heavy agent goals
• multi-query Brave search evidence gathering for research rounds
• new runtime and all-in-one configuration for deep research limits

v0.2.8-ce

SentientWave Automata Community Edition v0.2.8-ce adds two major capabilities: a new Cerebras LLM provider and a company-wide constitution and governance system enforced across agent reasoning.

Highlights

Cerebras LLM provider

Automata now supports Cerebras as a first-class LLM provider. Administrators can configure Cerebras through the existing LLM provider admin UI, use Cerebras-specific base URL and model defaults, and route agent inference through the Cerebras chat completions API while preserving structured error details for observability and trace storage.

Company constitution and governance

This release adds a new governance subsystem backed by PostgreSQL for published laws, proposals, votes, governance roles, elector snapshots, and immutable constitution snapshots. The constitution is no longer just documentation. It becomes an enforced system layer that is injected into new agent runs, planning paths, and related LLM calls.

Matrix voting workflow

Governance changes are proposed and voted on through a dedicated Matrix governance room. Proposal opening, voting, eligibility checks, and resolution now flow through Automata’s governance boundary with durable records stored in the database. Approved proposals publish a new immutable constitution snapshot instead of editing active laws in place.

Admin UI for constitution management

The Phoenix admin UI now includes a Constitution section with pages for law overview, proposal detail, proposal creation, governance roles, and role assignments. This gives administrators and organizations a visible control surface for managing the company constitution and governance process.

Included changes

• add Cerebras LLM provider support in backend and admin UI
• add governance schemas for laws, proposals, votes, roles, role assignments, elector snapshots, and constitution snapshots
• add governance Matrix room command parsing and dispatch for proposal and vote handling
• enforce constitution snapshot metadata in agent runtime and LLM trace context
• add admin UI pages for constitution overview, laws, proposals, and roles
• include governance room support in provisioning and runtime room detection

Validation

• mix compile --warnings-as-errors
• local all-in-one upgrade path validated during rollout testing

v0.2.7-ce

SentientWave Automata Community Edition v0.2.7-ce expands the platform in two important directions: richer operational administration and broader model connectivity.

This release introduces a new directory admin console for managing human, agent, and service identities from one place. Administrators can now inspect and update directory records, rotate credentials, review Matrix account state, and work with agent-specific runtime controls in a more complete workflow. For agent identities, the admin experience now includes runtime settings, online or offline operational status, designated tool access controls, and scheduled task management.

Scheduled tasks are now a first-class part of the agent operating model. Admins can create and manage recurring agent-owned work, including prompt-driven runs and room message tasks, using hourly, daily, and weekly schedules. This makes it much easier to keep Automata active in ongoing operations without requiring manual prompting every time.

On the LLM side, Automata now supports Anthropic as a provider alongside the existing provider abstraction. Anthropic models can be configured through the admin UI using the same provider management workflow already used for the rest of the system. The integration maps Automata's internal message structure into Anthropic's Messages API shape and preserves structured provider errors for better observability and future tuning workflows.

This release also improves day-to-day operator confidence because the all-in-one hot-upgrade path was validated against the current release state. We verified that upgrading through deploy/all-in-one/bin/upgrade.sh completes successfully, restarts Automata cleanly, and keeps Matrix online during the process.

Important areas included in this release:

• directory admin UI for human, agent, and service users
• agent runtime detail views and operational status reporting
• designated tool override controls for agents
• admin-managed scheduled tasks for recurring agent work
• Anthropic LLM provider support in backend and admin configuration UI

Validation performed for this release included successful compilation with warnings treated as errors and successful all-in-one upgrade testing against the running local container.

v0.2.6-ce

SentientWave Automata Community Edition v0.2.6-ce is now available.

This release brings a broader step forward in making multi-agent collaboration durable, inspectable, and governable.

A major part of this release is the new global skill catalog backed by PostgreSQL. Skills are now managed as organization-wide reusable records instead of isolated per-agent entries, and administrators can designate or roll them back on specific agents while preserving assignment history. Active designated skills now participate directly in runtime execution because their markdown is injected into model requests as system-level instructions before normal context and memory.

This release also strengthens observability around model behavior. LLM provider requests and responses are now persisted in the database in a provider-independent tracing layer, including requester identity, room or private scope, IP address when available, response payloads, and error payloads. On top of that, the Phoenix admin UI now includes searchable and filterable LLM trace pages so administrators can inspect model activity and troubleshoot inference behavior from the web console.

On the admin UI side, this release adds the first complete workflow for skill management in Phoenix. Administrators can browse a searchable skill catalog, filter by enabled state, inspect parsed tool hints, create and edit skills, designate them to agents, and roll back active designations from a dedicated skill detail page. The README has also been updated with new admin UI screenshots so new users can understand the product experience faster.

Highlights:

• database-backed global skill catalog
• per-agent skill designation and rollback history
• legacy skill backfill into the new schema
• runtime injection of designated skill instructions into LLM requests
• persisted LLM trace storage for future analysis and tuning
• searchable admin UI for LLM trace introspection
• new Phoenix admin UI pages for skill catalog management
• README updated with new admin UI screenshots
• VERSION bumped to v0.2.6-ce

Included work in this release line:

• persistent LLM trace storage
• admin UI trace explorer for LLM requests and responses
• global skill catalog and agent designation workflow
• README screenshot update
• release preparation for v0.2.6-ce

Validation:

• mix format
• mix compile --warnings-as-errors

Known local environment blocker:

• mix test and mix precommit remain blocked in the current local environment by PostgreSQL test database authentication:
  • no pg_hba.conf entry for host "192.168.127.1", user "postgres", database "postgres", no encryption

v0.2.5-ce

SentientWave Automata v0.2.5-ce

Community Edition v0.2.5-ce continues the work of making Automata feel like a real operator product instead of a collection of admin screens.

Highlights

Better admin management flows

LLM providers and tools now have cleaner management UX.

List pages are focused on overview and actions, while create and edit flows live on dedicated pages. This reduces clutter and makes the admin experience easier to scan and operate.

Full internal theme support

The admin UI now includes a built in theme switch with auto, light, and dark modes.

This is not just a toggle. The shared admin visual system was updated so the interface actually supports all three modes across the shell, cards, tables, forms, alerts, buttons, and login screen.

Password generator update

Password generation now includes the digit 1 in the allowed character set.

Version update

VERSION has been bumped to v0.2.5-ce.

Why this release matters

This release improves day to day operator usability in practical ways.

The admin UI is moving toward a clearer mental model with proper list pages, dedicated detail pages, and appearance controls that feel native to the product. That makes Automata easier to manage for real teams running local or self hosted deployments.

Included pull requests

• #8 Release v0.2.5-ce from dev to main
• #7 Include digit 1 in generated passwords
• #6 Add full light dark auto theme support to admin UI
• #5 Represent tools and LLM providers as admin tables with dedicated pages

Validation

• mix compile

Community Edition note

-ce denotes Community Edition.