This PR makes @cloudflare/think, agents, @cloudflare/ai-chat, and @cloudflare/codemode support both AI SDK v6 and v7, rather than the v7-only cutover in #1922. It stacks on #1922 (base branch fix/jul9-ai-sdk-v7) and is offered as an alternative: take #1922 for a clean v7-only move, or this PR to keep v6 working too.

Why

• The v7-only migration forces every consumer of these packages to upgrade their own ai and @ai-sdk/* majors in lockstep. @cloudflare/ai-chat and @cloudflare/think declare ai as a required peer, so on npm 7+ a consumer staying on v6 hits an ERESOLVE install failure, not a warning.
• v7 keeps the v6 option names as aliases (system, stepCountIs, experimental_telemetry, onStepFinish, experimental_onToolCallFinish), so calling the SDK through the shared names lets one code path serve both majors without a compile-time fork.
• The alternative was maintaining two builds or a version-conditional type layer. That was rejected: the only genuine runtime divergences are the tool-finish event shape, the UI stream helper, and one tool-result content part, all of which are cheaper to normalize at the boundary than to fork.
• Full type parity across both majors is not attempted. The shipped .d.ts is compiled against one major; consumers get correct types for whichever ai they install, and the runtime is correct for both.

Public API Surface

No exported symbols are added or removed. The change is to peer acceptance ranges only.

Code Changes

• @cloudflare/think calls streamText through the names present in both majors: system, stepCountIs, experimental_telemetry, onStepFinish, and experimental_onToolCallFinish. In v7 the last resolves internally to onToolExecutionEnd and fires once, so there is no double dispatch.
• normalizeToolFinishEvent collapses the two tool-execution-finished event shapes ({ success, output, error, durationMs, stepNumber } in v6, { toolOutput, toolExecutionMs } in v7) into one ToolCallResultContext so afterToolCall and extensions are version-agnostic.
• The UI message stream uses the result's toUIMessageStream() method, which exists in both majors. The standalone toUIMessageStream({ stream }) helper and result.stream are v7-only.
• The workspace read tool emits the file-data model-output content part, accepted by both majors. v7's newer { type: "file", data: { type: "data", data } } shape does not exist in v6.
• A test helper called .getReader() on toUIMessageStream()'s result, which became a pure async iterable when readableStreamToAsyncIterable was introduced in the v7 migration. The resulting throw masked the intended error and failed 17 error-handling tests. It now consumes the async iterator.
• Dropped the telemetry test that asserted on v7's telemetry.integrations callback API, which does not exist in v6, to avoid freezing v7-only behaviour.
• Added .github/workflows/ai-sdk-compat.yml and scripts/use-ai-sdk-major.mjs: a matrix that installs each ai major and runs @cloudflare/think's typecheck and workers tests under both.

Compatibility

• Consumers can stay on ai@6 / @ai-sdk/*@3 or move to ai@7 / @ai-sdk/*@4. Bumping these packages no longer forces an AI SDK major upgrade.
• workers-ai-provider (Think's default model provider) has no v7 release yet, so a consumer on ai@6 who relies on Think's built-in default model may hit a provider-version mismatch. Passing an explicit LanguageModel avoids this.
• agent-think (private, not published) uses the workers-ai-provider v7 preview from Add support for AI SDK 7 ai#601 and must be repointed at the release before merge, per the warning above.