Add a dedicated Ollama AI provider (ai/ollama/) that auto-detects
local vs cloud mode based on the base URL:
- Local Ollama: native /api/chat endpoint with NDJSON streaming
- Ollama Cloud: OpenAI-compatible /v1/chat/completions with SSE streaming
Both modes support tool calls with a multi-round execution loop.
Add agent.BaseURL option so agents can point at non-default LLM
endpoints (e.g. local Ollama, proxies). Wire it through micro.AgentBaseURL
at the top level.
Include a complete example (examples/agent-ollama/) demonstrating a
knowledge-base service with auto-discovered tools, a custom time tool,
streaming, and env-var configuration for local vs cloud.
Closes#3632
Fixes#2988. Brings 'golangci-lint run ./...' to zero issues (was ~373):
- errcheck: explicitly ignore fire-and-forget calls with '_ =' (and a small
errcheck.exclude-functions list for response writes — json Encoder.Encode,
http ResponseWriter.Write, fmt.Fprint*); genuine cases handled.
- unused: remove dead code (unexported decls and dead test helpers) and the
imports they orphaned.
- staticcheck: ST1005 error strings, ST1016 receiver names, S1000/S1017/S1019/
S1023 simplifications, SA4004/SA4006/SA4010 dead code, SA1021 net.IP.Equal,
SA6002 (store *[]byte in sync.Pool).
- govet: fix a context leak (lostcancel) in internal/util/mdns and move
t.Fatal/Fatalf out of goroutines (testinggoroutine) in tests.
- ineffassign, unconvert: mechanical fixes.
CI: the Lint workflow now runs a blocking full-tree 'golangci-lint run' on
pushes and PRs (dropped only-new-issues now that the tree is clean).
Verified: go build, go vet, test compilation, and unit tests for the
behaviourally-touched packages all pass.
Claude-Session: https://claude.ai/code/session_01CmdEY7pYmV5zzwCjNJ4ykL
Co-authored-by: Claude <noreply@anthropic.com>
Closes the remaining ask in #2980 without adding a parallel callback API.
ToolResult.Refused tags a guardrail block with a reason (ai.RefusedLoop /
RefusedMaxSteps / RefusedApproval) so a wrapper can switch on it instead of
parsing the message. ai.RunInfo (RunID, ParentID, Agent) rides on the
context passed to the tool handler, giving wrappers run correlation and
delegation lineage. Before/after/retry/failure were already covered by
AgentWrapTool; this adds the metadata. Docs + tests included.
Co-authored-by: Claude <noreply@anthropic.com>
* test(harness): read agent plan from the scoped store
The store-scoping change moved an agent's plan from the default table
key agent/{name}/plan to its own table (database "agent", table {name},
key "plan"). The plan-delegate harness tests still read the old key and
failed with 'not found'; read through store.Scope(mem, "agent", name)
like the agent does.
* docs: orient agents-first across README, landing, and docs overview
Lead with agents (then services and flows), surface MCP + A2A as the
interop story, and frame agents as services. Landing hero and feature
grid reordered agents-first with an A2A gateway card.
* v6: module path go-micro.dev/v6, TLS secure by default, NewService
Cut v6. Three breaking changes, bundled so the major bump is paid once:
- Module path go-micro.dev/v5 -> go-micro.dev/v6 across all imports + go.mod.
- TLS verification on by default (was off). MICRO_TLS_SECURE removed;
MICRO_TLS_INSECURE=true opts out for self-signed/dev.
- micro.NewService(name, opts...) is the canonical service constructor,
symmetric with NewAgent/NewFlow; micro.New kept as a deprecated alias;
the old name-less NewService(opts...) removed. Generators emit NewService.
Also ports the JWT auth token provider in-module (go-micro.dev/v6/auth/jwt/token
on golang-jwt/jwt/v5), dropping the v5-pinned github.com/micro/plugins/v5/auth/jwt
and the deprecated dgrijalva/jwt-go.
Docs/README/landing updated to v6 and @latest; v5->v6 migration guide added;
CHANGELOG cut as [6.0.0]. Blog posts left at their historical versions.
---------
Co-authored-by: Claude <noreply@anthropic.com>
Refactor the A2A handler into a reusable dispatcher + Invoke seam and
expose NewAgentHandler(card, invoke) + Card(). An agent now serves its
own A2A endpoint with AgentA2A(addr) / WithA2A — handling tasks
in-process (no RPC hop, no separate gateway). The gateway and embedded
agent share the same handler; the only difference is RPC vs in-process
invocation. Docs, README, and changelog cover both deployment modes.
Co-authored-by: Claude <noreply@anthropic.com>
* docs: design note for flow steps + Checkpoint durable execution
* docs: fold in durable-execution decisions (State struct, single Step, run retention, retry)
* docs: rename State.Payload to State.Data
* feat(flow): ordered steps + Checkpoint durable execution
A flow can now be an ordered list of steps (a task with stages) instead
of a single LLM turn. State carries typed Data plus a Stage marker; each
step is checkpointed before and after via a pluggable Checkpoint
(store-backed by default), so a run survives a crash and resumes where it
stopped without re-running completed steps. Flow-level Retry with a
per-step override; runs retained for audit unless DeleteOnSuccess.
Step actions: Call (RPC), LLM (augmented turn), Dispatch (to an agent),
or any StepFunc. Single-step and agent-dispatch flows are unchanged.
* feat(flow): top-level re-exports + durable flow example
Expose the step/checkpoint API from the micro package (FlowSteps,
FlowStep, FlowState, FlowRetry, FlowWithCheckpoint, FlowCall/LLM/Dispatch,
Checkpoint, StoreCheckpoint) and add a runnable, key-free example
demonstrating crash + resume.
* docs: document durable flow steps (guide, README, CLI help)
* docs: blog post + changelog for durable workflows
* fix(flow): scope checkpoint keys by flow name (flow/{name}/runs/{id})
Run keys were flow/runs/{id} — a single global keyspace shared by every
flow on the default store. Namespace them by flow name so each flow's
state is kept apart. StoreCheckpoint now takes a scope argument (the flow
passes its name by default).
* feat(store): Scope handle; scope agent and flow state by name
Add store.Scope(s, database, table) — a store handle that confines every
operation to a database/table without mutating the shared store, so
co-located components don't clobber each other's table (the failure mode
of the global Init(Table(...)) approach).
Use it to keep each agent's memory and plan in its own table
(agent/{name}) and each flow's runs in its own (flow/{name}), instead of
one global table partitioned only by key prefix. Services already scope
by service name.
* feat: consistent state model — service store scoping, flow registry, list/history CLI
- service: scope store via store.Scope (database service / table name),
retiring the Init(store.Table(name)) global-mutation hack; bridge the
default store so handlers using store.DefaultStore stay isolated.
- flow: register in the registry as type=flow while running (with trigger
and step count), deregister on Stop. Live discovery, like agents.
- cli: micro flow list (registry), micro flow runs <name> (durable store),
micro agent history <name> (durable store). list = running, runs/history
= durable, mirroring the service model.
* test: mini-universe end-to-end harness + scheduled GitHub Action
internal/harness/universe boots a small but real go-micro world — four
services, a durable checkout flow that crashes at payment and resumes,
and a guardrailed agent with a tool wrapper reached over RPC — drives the
scenario, asserts the end state (10 checks), and shuts down. Everything
is real except the LLM (mocked), so it's deterministic and needs no key;
-provider anthropic runs it live. Exits non-zero on failure, so it's an
end-to-end test, not just a demo.
Adds .github/workflows/universe.yml (push/PR/daily/dispatch) running the
universe + existing harnesses on the mock provider, plus an opt-in job
that runs live when ANTHROPIC_API_KEY is set. 'make harness' runs them
locally.
* ci: run the live universe job against AtlasCloud (ATLASCLOUD_API_KEY)
* ci: run the live universe job only on schedule or manual dispatch
The deterministic mock job still runs on push/PR/daily; the live
(AtlasCloud) job runs daily and on manual workflow_dispatch only, so
changes don't burn API credits on every PR but can still be checked
against a real model on demand.
---------
Co-authored-by: Claude <noreply@anthropic.com>
Add LoopLimit: refuse a tool call repeated with identical arguments in
one Ask, with a self-heal message so the model changes approach. Catches
the no-progress loop that MaxSteps (count) and the gateway circuit
breaker (failures) miss. Enforced at the same tool-handler choke point as
MaxSteps/ApproveTool; on by default (lenient 3); AgentLoopLimit(0) to
disable. Tests cover repeats, distinct calls, disabled, and default-on.
Docs: new Agent Guardrails guide (MaxSteps/LoopLimit/ApproveTool, the
ApproveTool integration seam for external policy engines, and the
gateway's RateLimit/CircuitBreaker), nav + README + AGENT_DESIGN updates,
and blog/23 'Agent Guardrails'.
Co-authored-by: Claude <noreply@anthropic.com>
* docs: add 'become a sponsor' call-to-action linking to Discord
Now that there are a couple of sponsors, invite more: a short CTA under
the Sponsors section in the README and on the landing page, pointing to
the Discord to get in touch.
* fix(health): remove duplicate RegistryCheck declaration
Two PRs (#2957 and #2958) each added a RegistryCheck to the health
package, leaving the package uncompilable on master (RegistryCheck
redeclared: health/registry.go vs health/health.go). Keep the
health.go implementation — it honors the check's context timeout so a
hung registry (e.g. an unreachable etcd) reports down instead of
blocking the probe — and remove the duplicate registry.go and its test.
registry_check_test.go already covers healthy/down/nil/timeout/not-ready.
* feat(agent): pluggable memory and custom tools
Make agents compose the way services do — pluggable pieces with working
defaults — by adding the two abstractions an agent needs beyond the model:
- Memory: a pluggable interface for conversation memory. The default is
store-backed and durable across restarts (the previous hardcoded
behavior, now behind an interface); supply your own with WithMemory
(in-memory, database, semantic store). NewMemory / NewInMemory provided.
- Custom tools: WithTool registers any function as a tool the agent can
call, so agents are no longer limited to orchestrating RPC services.
Both exposed at the micro package (AgentMemory, AgentTool, NewMemory,
NewInMemory). Behavior-preserving refactor of the agent's history into
the default Memory; tests cover persistence, in-memory, clear, custom
tool dispatch and errors. README + AGENT_DESIGN document the pluggable
composition (model / memory / tools / guardrails).
* blog: 'Doubling Down on Agents' (#20)
The vision post for making agents a first-class framework the way
services were: opinionated, batteries-included, pluggable. Frames an
agent as a composition of model + memory + tools + guardrails with
working defaults; introduces the new pluggable memory and custom tools;
makes the microagents argument (an agent for everything, distributed
like microservices); and lays out the three primitives — services,
agents, workflows — as one substrate, with an honest list of the gaps
still to fill (knowledge/retrieval, streaming, explicit loop).
---------
Co-authored-by: Claude <noreply@anthropic.com>
* docs: map go-micro onto Anthropic's workflows-vs-agents taxonomy
- new guide 'Agents and Workflows': adopts Anthropic's Building Effective
Agents vocabulary — workflow (predefined path) = flow, agent (dynamic
self-direction) = agent — maps the augmented-LLM building block and the
five workflow patterns onto go-micro, and shows routing (chat router)
and orchestrator-workers (conductor + plan/delegate) are already native.
- flow package doc reframed as a workflow (predefined path) per the same
taxonomy, with guidance on flow vs agent.
- nav + README link the new guide.
* feat: agent guardrails — step limit and tool approval hook
Anthropic's Building Effective Agents stresses stopping conditions and
human-in-the-loop checkpoints for autonomous agents. Add both as plain
options enforced at the tool-handler choke point — no provider changes,
no new abstraction:
- MaxSteps(n): bound tool executions per Ask; beyond the limit, actions
are refused and the model is told to stop and summarize.
- ApproveTool(fn): gate each action before it runs; returning false
blocks it and surfaces the reason to the model. The internal plan tool
is never gated.
Exposed at the micro package (AgentMaxSteps, AgentApproveTool, ApproveFunc).
Tests cover the limit, blocking, and that plan is not gated. Guardrails
section of the agents-and-workflows guide updated from 'active work' to
documented options.
* feat: flow can dispatch to an agent (flow triggers, agent reasons)
Unify the engine without collapsing the workflow/agent distinction. A
Flow with Agent set hands each event's rendered prompt to a named
registered agent over RPC (Agent.Chat) instead of running its own LLM
step — so the workflow stays the deterministic trigger and the agent is
the reasoning engine, with its plan, delegate, memory, and guardrails.
A plain flow is unchanged (single augmented-LLM step).
- flow.Agent(name) / micro.FlowAgent(name); flow stores the client and
skips model setup when dispatching.
- test: dispatch routes to comms.Agent.Chat with the rendered prompt and
records the reply.
- guide: 'Flow triggers, Agent reasons' section.
---------
Co-authored-by: Claude <noreply@anthropic.com>
Asim Aslam