[dependabot]

Bumps ruby_llm from 1.16.0 to 2.1.1.

Release notes

Sourced from ruby_llm's releases.

RubyLLM 2.1.1 fixes fresh installs that today's faraday-net_http 3.5.0 release broke, connects to MCP servers that predate the 2026-07-28 revision, and keeps batch embeddings paired with the right inputs.

bundle add ruby_llm --version 2.1.1

Already on 2.1? Run bundle update ruby_llm. No schema or code changes are needed.

Fixed

• Install RubyLLM with the latest faraday-net_http. faraday-net_http 3.5.0 falls outside the range every Faraday from 2.0.1 to 2.14.4 accepts, so a fresh bundle install resolved Faraday 2.0.0, which loads no default adapter, and every request failed with :net_http is not registered on Faraday::Adapter. RubyLLM now excludes Faraday 2.0.0. By @​crmne; thanks @​juneira. (ef2bebdb)
• Connect to MCP servers that predate the 2026-07-28 revision. A server that rejects the revision with -32022 now gets the initialize handshake whether it lists the legacy revisions it speaks, as Atlassian's does, or lists none, as Datadog's does. By @​ebeigarts, @​juneira, and @​crmne. (#1056, #1057, #1053, #1055)
• Pair batch embeddings with the right inputs. OpenAI-compatible and Mistral batch results now place each vector by the position the provider reports, and mark a result with duplicate, missing, or out-of-range positions as failed instead of pairing its vectors with the wrong texts. By @​D41910. (#1051, #1052)
• Keep the Bedrock inference profile you choose. An id that already names a geography, such as eu.anthropic.claude-sonnet-5-5, is no longer rewritten to another region's profile, so requests stay where your IAM policy allows them. By @​s-andringa. (#1050)
• Keep chat replies when a usage row cannot be linked. A persisted chat saves the reply and logs the failed link, instead of rolling back an answer the provider already billed. By @​yorzi. (#1048, #1049)
• Upload converted attachments again after a provider deletes them. The chat forgets the stale upload and retries, as it already did for plain attachments. By @​yorzi. (#1046, #1047)

Thanks

Thanks to @​yorzi, @​s-andringa, @​D41910, @​ebeigarts, and @​juneira for the fixes and reports in this release.

Full changelog: crmne/ruby_llm@v2.1.0...v2.1.1

RubyLLM 2.1.0 adds a built-in MCP client, typed judgments, agent evaluations, and OpenTelemetry tracing. Conversations use less memory, reuse connections, and keep working when you switch providers.

bundle add ruby_llm --version 2.1.0

Upgrading from 2.0? Ruby 3.2 or later is required. Run the Rails upgrade generator and read the upgrade guide before deploying. Applications on 1.x must finish the 2.0 upgrade first. See What's New in 2.1 for examples.

rubyllm-tour.mp4

Take the three-minute tour in full quality (1080p60).

New

• Connect agents to MCP servers from Ruby. Class-based servers expose tools, resources, and prompts through the chat API, with OAuth, encrypted Rails credentials, approvals, input requests, change notifications, and resumable background tasks. MCP Apps retain the results their UIs need. By @​crmne. (#959, #965, #974)
• Ask typed questions about your data. RubyLLM::Judge returns probabilities, choices, and scores through TypeSafe Jev, OpenAI Decisions, and local decision models through Ollama. TypeSafe and Hetzner bring the built-in provider count to nineteen. By @​crmne and @​kieranklaassen. (#1008, acecd62f, 16ad59b0, 689d6c7a)
• Evaluate agents against reference answers. RubyLLM::Evaluation runs datasets, grades correctness or your own criteria, supports Ruby assertions and RSpec/Minitest integration, and reports verdicts, progress, tokens, and cost. By @​crmne. (10eb5885, 50df0d14, e2513142, d259e1d6)
• Trace model calls, tools, and workflows with OpenTelemetry. Enable the optional integration with your application's SDK and exporters. Spans follow the GenAI conventions and omit prompts and responses. By @​crmne. (45970ee5)
• Give models large tool collections without sending every definition. defer: true enables native tool search on Anthropic and OpenAI Responses. Other protocols receive ordinary tool definitions. Local and MCP tools share after_tool_progress for showing work as it happens. By @​crmne. (#980, a76ece81, 9a977de3)
• Spend less time and memory inside RubyLLM. Streaming parses and appends incrementally, model lookups use an index, and replies stop retaining their requests. Persistent Faraday adapters reuse connections; Vertex AI shares credentials; Rails avoids unnecessary queries and downloads and reuses stored provider uploads. See the measured comparisons. By @​crmne and @​kalicki. (#981, 15b48ed0, d5ea078e, b7417796)
• Convert files a protocol cannot send. convert_unsupported_attachments lets your application supply a readable replacement once per file while preserving the original transcript. resolution: controls supported image detail, including :original. By @​crmne, @​guizaols, and @​whatthewhat. (#989, #1005, d14b1f7a, 020d6876)
• Share prompt partials and configure agents per tenant. Prompt templates render partials with locals. Runtime agent contexts can use inputs and persisted records to select credentials and endpoints. RubyLLM's supporting records can share your application's secondary database. By @​kryzhovnik, @​mikemikimike, and @​crmne; thanks @​BigBlue79. (#956, #903, fc66b9f0)
• Account for operations outside persisted chats. One-shot operations and completed video/research jobs write Rails usage rows, with owner attribution. Provider tool counts share names and persist with usage; xAI and OpenRouter video results retain reported costs. By @​crmne. (073ab733, 9775ee90, 48278dfa, 26fe4ca2)
• Use Azure deployment names with model pricing. Deployment mappings retain the underlying model's pricing and limits. Perplexity chat moves to the Agent API, with compatibility warnings for old Sonar names. By @​guizaols, @​Halvanhelv, and @​crmne. (#987, #985, #994, #995)
• Give each cache boundary its own lifetime. cache_until_here(ttl: "1h") keeps a stable prefix cached for an hour while the rest of the chat uses the with_caching default, on Anthropic, OpenRouter, and Bedrock Converse. Rails keeps the lifetime in a new cache_ttl message column. By @​crmne; thanks @​nwumnn. (#1041)

... (truncated)

Changelog

Sourced from ruby_llm's changelog.

RubyLLM 2.1.1 fixes fresh installs that today's faraday-net_http 3.5.0 release broke, connects to MCP servers that predate the 2026-07-28 revision, and keeps batch embeddings paired with the right inputs.

bundle add ruby_llm --version 2.1.1

Already on 2.1? Run bundle update ruby_llm. No schema or code changes are needed.

Fixed

• Install RubyLLM with the latest faraday-net_http. faraday-net_http 3.5.0 falls outside the range every Faraday from 2.0.1 to 2.14.4 accepts, so a fresh bundle install resolved Faraday 2.0.0, which loads no default adapter, and every request failed with :net_http is not registered on Faraday::Adapter. RubyLLM now excludes Faraday 2.0.0. By @​crmne; thanks @​juneira. (ef2bebdb)
• Connect to MCP servers that predate the 2026-07-28 revision. A server that rejects the revision with -32022 now gets the initialize handshake whether it lists the legacy revisions it speaks, as Atlassian's does, or lists none, as Datadog's does. By @​ebeigarts, @​juneira, and @​crmne. (#1056, #1057, #1053, #1055)
• Pair batch embeddings with the right inputs. OpenAI-compatible and Mistral batch results now place each vector by the position the provider reports, and mark a result with duplicate, missing, or out-of-range positions as failed instead of pairing its vectors with the wrong texts. By @​D41910. (#1051, #1052)
• Keep the Bedrock inference profile you choose. An id that already names a geography, such as eu.anthropic.claude-sonnet-5-5, is no longer rewritten to another region's profile, so requests stay where your IAM policy allows them. By @​s-andringa. (#1050)
• Keep chat replies when a usage row cannot be linked. A persisted chat saves the reply and logs the failed link, instead of rolling back an answer the provider already billed. By @​yorzi. (#1048, #1049)
• Upload converted attachments again after a provider deletes them. The chat forgets the stale upload and retries, as it already did for plain attachments. By @​yorzi. (#1046, #1047)

Thanks

Thanks to @​yorzi, @​s-andringa, @​D41910, @​ebeigarts, and @​juneira for the fixes and reports in this release.

Full changelog: crmne/ruby_llm@v2.1.0...v2.1.1

Commits

• a40182d Prepare RubyLLM 2.1.1 for release
• e5a7a42 MCP: shake hands when a version reject lists nothing
• a8c17eb Check the Faraday 2.0.0 exclusion under the v1 job too
• ef2bebd Exclude Faraday 2.0.0, which has no default adapter
• e9dcd62 MCP: fall back to the handshake when a version reject lists a legacy revision...
• fa28fe3 Check Mistral batch embedding row positions before pairing vectors with input...
• 5556592 Check embedding batch row positions before pairing vectors with inputs (#1051)
• a67da2c Keep Bedrock ids that already name an inference profile (#1050)
• 93e2179 Keep chat replies when a usage row cannot be linked (#1049)
• 661f3d8 Restore uploads for converted attachments (#1047)
• Additional commits viewable in compare view

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

---

Dependabot commands and options

You can trigger Dependabot actions by commenting on this PR:

• @dependabot rebase will rebase this PR
• @dependabot recreate will recreate this PR, overwriting any edits that have been made to it
• @dependabot show <dependency name> ignore conditions will show all of the ignore conditions of the specified dependency
• @dependabot ignore this major version will close this PR and stop Dependabot creating any more for this major version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this minor version will close this PR and stop Dependabot creating any more for this minor version (unless you reopen the PR or upgrade to it yourself)
• @dependabot ignore this dependency will close this PR and stop Dependabot creating any more for this dependency (unless you reopen the PR or upgrade to it yourself)

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	⚠️ Failed 2026-10-19T21:43:47.830388Z	fbffc61	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.