docs(docker-agent): add new model providers and config fields#25478
docs(docker-agent): add new model providers and config fields#25478aheritier wants to merge 4 commits into
Conversation
…76,3384,3385,3337,3389 - add new model providers Add documentation for twelve new built-in provider aliases: - Baseten (docker#3341) - OVHcloud (docker#3343) - Groq (docker#3358) - DeepSeek (docker#3361) - Cerebras (docker#3368) - Fireworks AI (docker#3369) - Together AI (docker#3376) - Moonshot AI / Kimi (docker#3384) - Vercel AI Gateway (docker#3385) - OpenRouter (docker#3337) - Cloudflare Workers AI (docker#3389) - Cloudflare AI Gateway (docker#3389)
… model option Document the bypass_models_gateway field, which lets a specific model connect directly to its provider even when a models gateway is configured.
Document the compaction_model field, which lets agents delegate session compaction (summary generation) to a cheaper, faster model.
Document the tags field on the metadata section, used for agent categorization and discovery when publishing to OCI registries.
✅ Deploy Preview for docsdocker ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
docker-agent
left a comment
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟢 APPROVE
The changes correctly add documentation for twelve new model provider aliases and three new configuration fields (compaction_model, bypass_models_gateway, and tags). No vendored files were touched, no redirects are needed, and no style guide violations (hedge words, AI-isms, marketing language) were found in the new content. The compaction_model NOTE callout is logically consistent and accurate.
docker-agent
left a comment
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟢 APPROVE
The added documentation for compaction_model, bypass_models_gateway, version, and tags is well-structured, accurate, and proportionate in scope. Three minor style issues were found — all low severity and non-blocking.
|
|
||
| > [!NOTE] | ||
| > If the compaction model has a smaller context window than the primary model, | ||
| > Docker Agent triggers compaction against the smaller window, so the summary |
There was a problem hiding this comment.
[LOW] NOTE callout may confuse readers about compaction behavior
The NOTE describes two scenarios without clearly distinguishing them:
- "If the compaction model has a smaller context window … Docker Agent triggers compaction against the smaller window, so the summary call can always ingest the full conversation." (Describing behavior when a smaller-window model is used — compaction fires early based on the smaller window, so it works.)
- "Pair the primary with a compaction model whose context window is at least as large to keep the proactive trigger aligned with the primary's window." (Recommending against using a smaller-window model.)
The two sentences describe conflicting scenarios without framing one as "here is what happens" and the other as "and here is why you should avoid it." Consider restructuring:
If the compaction model has a smaller context window than the primary model, Docker Agent uses that smaller window to trigger compaction proactively — the summary call can still ingest the full conversation, but compaction fires earlier than expected. Use a compaction model whose context window is at least as large as the primary's to keep the trigger aligned with the primary's window.
| own model to summarize the conversation. That summary call ingests the entire | ||
| conversation and is the slowest, most expensive call in a session. | ||
|
|
||
| You can point `compaction_model` at a smaller, faster model to make compaction |
There was a problem hiding this comment.
[LOW] Informal phrasing "point compaction_model at" — prefer "set compaction_model to"
"You can point compaction_model at a smaller, faster model" uses the informal idiom "point X at Y". The Docker style guide prefers direct, precise language. Consider:
Set
compaction_modelto a smaller, faster model to reduce compaction cost without changing the model that runs the conversation:
| instruction: You are a helpful assistant. | ||
| ``` | ||
|
|
||
| A bypassed model authenticates with its own provider credentials |
There was a problem hiding this comment.
[LOW] Parentheses in prose — consider restructuring
The Docker style guide recommends avoiding parentheses in technical documentation as they reduce readability. The current sentence:
A bypassed model authenticates with its own provider credentials (
OPENAI_API_KEY,ANTHROPIC_API_KEY, or the explicittoken_key) rather than the gateway's token.
Could be rewritten without parentheses:
A bypassed model authenticates directly with its provider using
OPENAI_API_KEY,ANTHROPIC_API_KEY, or the explicittoken_key— not the gateway's token.
2 participants
Syncs Docker Agent documentation with changes merged into docker/docker-agent main in the last 36 hours.
New model providers
Adds documentation for twelve new built-in provider aliases:
Baseten, Cerebras, Cloudflare AI Gateway, Cloudflare Workers AI, DeepSeek,
Fireworks AI, Groq, Moonshot AI (Kimi), OpenRouter, OVHcloud, Together AI,
and Vercel AI Gateway.
Source PRs: #3341, #3343, #3358, #3361, #3368, #3369, #3376, #3384, #3385,
#3337, #3389 in docker/docker-agent.
New model configuration fields
compaction_model: delegate session compaction to a cheaper model(feat(compaction): allow a dedicated model for session summary generation docker-agent#3335)
bypass_models_gateway: connect a specific model directly to itsprovider, bypassing any configured gateway (feat: add per-model bypass_models_gateway option docker-agent#3338)
New metadata field
tags: string array for agent categorization and discovery in OCIregistries (feat: add tags field to agent config metadata docker-agent#3390)