Skip to content

docs(caching): provider count seven → eight + guard it #210

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Skobeltsyn merged 1 commit into from
Jun 15, 2026
+3 −2
Merged

docs(caching): provider count seven → eight + guard it #210

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions docs/caching.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,7 +36,7 @@ val agent = agent<String, String>("ResearchBot") {
| **DeepSeek** | Automatic disk-based caching | Inherits the OpenAI-compatible request shape; no extra wiring needed. Cached-input tokens surface on `TokenUsage`. |
| **Ollama** | Engine-level KV-cache reuse (no wire-level control) | Hints degrade to no-op. Prefix stability (see below) is what makes the engine cache hit. |

The four rows above are the four **wire shapes**; `ModelProvider.entries` has seven values — Kimi, OpenRouter, and Perplexity extend the OpenAI adapter and inherit its row (automatic prefix caching where the upstream supports it). Fourth-party deployments that ride on top of one of these wire shapes are documented under [Under evaluation](#under-evaluation) below.
The four rows above are the four **wire shapes**; `ModelProvider.entries` has **eight** values — Kimi, OpenRouter, and Perplexity extend the OpenAI adapter and inherit its row (automatic prefix caching where the upstream supports it), and Gemini is its own wire shape (its prompt caching is not wired in this table — engine-side, no `cache_control`). Fourth-party deployments that ride on top of one of these wire shapes are documented under [Under evaluation](#under-evaluation) below.

### `CacheHint` model

Expand Down Expand Up @@ -106,7 +106,7 @@ These are the silent killers — `System.currentTimeMillis()` interpolated into

### Under evaluation

These backends are **not** first-party `ModelProvider` adapters (`ModelProvider.entries` has seven first-party values; these are not among them). Consumers who point the OpenAI adapter at one of these endpoints (via `openAiBaseUrl`) get OpenAI-compatible behavior, but the caching characteristics differ:
These backends are **not** first-party `ModelProvider` adapters (`ModelProvider.entries` has eight first-party values; these are not among them). Consumers who point the OpenAI adapter at one of these endpoints (via `openAiBaseUrl`) get OpenAI-compatible behavior, but the caching characteristics differ:

| Backend | Wire-shape compatibility | Caching note |
|---|---|---|
Expand Down
1 change: 1 addition & 0 deletions src/test/kotlin/agents_engine/core/DocsConsistencyTest.kt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -39,6 +39,7 @@ class DocsConsistencyTest {
val word = countWords[ModelProvider.entries.size]
val claims = listOf(
"docs/providers.md" to "`ModelProvider.entries` has **$word** values",
"docs/caching.md" to "`ModelProvider.entries` has **$word** values",
"docs/model-and-tools.md" to "$word providers ship today",
"docs/model-and-tools.md" to "All $word providers share the `ModelClient` interface",
"SECURITY.md" to "$word first-party providers",
Expand Down
Loading