* ai/minimax: complete provider surface (matrix, conformance, changelog)
Follow-up after merging the MiniMax provider (#3769), mirroring the Ollama
completeness pass (#3637):
- Add the `minimax` row to the AI provider capability matrix and blank-import
ai/minimax in provider_capabilities_test.go so the matrix stays enforced
against the registry.
- Add minimax to the stream-conformance allowlist (+ import) so its streaming
is actually exercised against the OpenAI-compatible SSE contract, not just
registered. It passes via the shared ai/internal/openaiapi path.
- Record the provider in CHANGELOG [Unreleased].
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CmdEY7pYmV5zzwCjNJ4ykL
* ai: update capabilities_test provider assertions for minimax
Adding the minimax blank-import to the shared ai_test binary (for stream
conformance) also registers it for TestRegisteredProviders / TestCapabilityRows
/ TestCapabilityMatrix in capabilities_test.go, which pin the exact provider
set. Update those assertions to include minimax. (Fixes the Unit Tests failure
my scoped `-run TestStreamProviders` check missed — go compiles all _test.go in
a package into one binary.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CmdEY7pYmV5zzwCjNJ4ykL
---------
Co-authored-by: Claude <noreply@anthropic.com>
Follow-up cleanup after merging the Ollama provider (#3636):
- Add the `ollama` row to the AI provider capability matrix in the provider
guide, and blank-import `ai/ollama` in provider_capabilities_test.go so the
matrix stays enforced against the registry (the provider registers a stream
but wasn't imported in that test, so its row went unchecked).
- README: bump "7 LLM providers" → 8 and list Ollama (local + cloud); add its
default model (`llama3.2`) to the model table.
- Fix a fictional model name shipped in the example and package doc:
`gemma4:31b-cloud` → `gpt-oss:120b`. gemma4 doesn't exist, and the `-cloud`
suffix is for cloud models proxied through a local Ollama, not the direct
ollama.com/v1 endpoint the example uses.
- Record the provider and the new agent.BaseURL/micro.AgentBaseURL option in
the CHANGELOG [Unreleased] section.
Claude-Session: https://claude.ai/code/session_01CmdEY7pYmV5zzwCjNJ4ykL
Co-authored-by: Claude <noreply@anthropic.com>
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
* atlascloud: env-selectable chat model; run conformance on a stronger model
The daily provider-conformance harness fails 4/5 harnesses on Atlas Cloud —
its default chat model answers agent/tool-use conformance prompts
conversationally instead of performing the task. Atlas is currently the only
provider with a key configured, so the whole live run is red.
Make the Atlas Cloud provider honor an ATLASCLOUD_MODEL env override (falling
back to the existing default), and set it in the harness workflow to a
stronger tool-use model (Qwen3, overridable via an Actions variable). No
change to the default for normal use.
* atlascloud: use minimaxai/minimax-m3 for conformance model
---------
Co-authored-by: Claude <noreply@anthropic.com>
Streaming is implemented now (v6.3.3); the old test expected
ErrStreamingUnsupported and failed. Replace it with a real SSE streaming
test (httptest) that also asserts stream_options.include_usage and the
final usage chunk.
* ai/atlascloud: surface token usage on streams
Request stream_options.include_usage and return the final usage chunk
as a Response with Usage set, so streaming callers can record usage.
* ai/openai: surface token usage on streams
Request stream_options.include_usage and return the final usage chunk
as a Response with Usage set, so streaming callers can record usage.
* Initial plan
* fix: remove extra blank line in ai/anthropic/anthropic.go (gofmt)
---------
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
* ai: add WithMaxTokens option
Let callers cap response length; providers send max_tokens when set
(anthropic keeps its 8192 default otherwise).
* ai: add WithMaxTokens option
Let callers cap response length; providers send max_tokens when set
(anthropic keeps its 8192 default otherwise).
* ai: add WithMaxTokens option
Let callers cap response length; providers send max_tokens when set
(anthropic keeps its 8192 default otherwise).
* ai: add WithMaxTokens option
Let callers cap response length; providers send max_tokens when set
(anthropic keeps its 8192 default otherwise).
* ai/atlascloud: thread Request.Messages into the request
Fold conversation history (req.Messages) between the system prompt and
the final user prompt so multi-turn context reaches the model.
Refs #3292
* ai/openai: thread Request.Messages into the request
Fold conversation history (req.Messages) between the system prompt and
the final user prompt so multi-turn context reaches the model.
Refs #3292
* ai/anthropic: thread Request.Messages into the request
Fold conversation history (req.Messages) between the system prompt and
the final user prompt so multi-turn context reaches the model.
Refs #3292
Follow-up to #3017. A Generate runs the whole tool-execution turn, so
auto-retrying it re-runs already-executed (possibly side-effecting) tool
calls. Default ModelMaxAttempts 3 -> 1: retries are now opt-in via
ModelRetry. The timeout stays as a safety net.
Also harden ai.GenerateWithRetry: always back off between retries
(exponential, capped at 30s, default 200ms if unset) so an opt-in retry
can't busy-loop the provider even with Backoff=0.
Verified: go build, go test -race ./agent/... ./ai/, golangci-lint.
Co-authored-by: Claude <noreply@anthropic.com>
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>
* docs: compare Go Micro with Google ADK in the comparison guide
Adds a 'vs Agent Frameworks (Google ADK)' section: ADK builds an agent,
Go Micro builds the distributed system the agent lives in (agents are
services in the mesh). Covers the category difference, a feature table,
when to choose each, and MCP/A2A interoperability.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CmdEY7pYmV5zzwCjNJ4ykL
* docs: replace ADK comparison slogan with concrete explanation
State plainly what each tool provides (ADK builds an agent process; Go Micro
builds the surrounding service mesh) instead of marketing phrasing.
* lint: apply golangci-lint autofixes; exclude ST1003 and demo errcheck
Mechanical, behaviour-preserving fixes applied by 'golangci-lint run --fix':
gofmt, misspell (US spelling), usestdlibvars (http.Method*/Status*), unconvert,
and the auto-fixable staticcheck simplifications (QF*, S1017/S1019/S1023/S1039).
Config: exclude ST1003 (remaining offenders are exported API renames, e.g.
web.Id, which would break compatibility) and skip errcheck for examples/ and
internal/harness/ (demo code where fire-and-forget is intentional).
Build and test compilation verified.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01CmdEY7pYmV5zzwCjNJ4ykL
* lint: WIP cleanup checkpoint (errcheck config + partial fixes)
Checkpoint of an in-progress golangci-lint cleanup (background pass). Builds
cleanly; lint is not yet zero. Follow-up commit will complete the cleanup and
switch CI to a blocking full-tree lint.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
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>
Codex