* feat(ai): add Cloudflare AI Gateway as a provider
Routes through Cloudflare's Unified API (`/compat`) for Workers AI and
Anthropic models, and through the provider-specific `/openai` subpath
for OpenAI models so reasoning models (gpt-5.x, o-series) can hit
`/v1/responses` natively. Once `/compat` adds Responses-API support,
the OpenAI subpath can be folded back in.
Catalog layout:
workers-ai/@cf/... -> openai-completions, gateway/.../compat
anthropic/... -> openai-completions, gateway/.../compat
<native-id> -> openai-responses, gateway/.../openai
(gpt-5.1, claude-... no, sorry: gpt-5.x and o-series only;
prefix stripped because the OpenAI SDK posts native ids)
Touches:
packages/ai/src/types.ts add cloudflare-ai-gateway to KnownProvider
packages/ai/src/env-api-keys.ts map to CLOUDFLARE_API_KEY
packages/ai/src/providers/cloudflare.ts add CLOUDFLARE_AI_GATEWAY_COMPAT_BASE_URL
and CLOUDFLARE_AI_GATEWAY_OPENAI_BASE_URL
packages/ai/src/providers/openai-responses.ts one-line dispatch through resolveCloudflareBaseUrl
(matches what openai-completions.ts already does)
packages/ai/scripts/generate-models.ts branch openai/* vs workers-ai/anthropic/*
packages/ai/src/models.generated.ts spliced 34 entries
packages/ai/test/stream.test.ts 3 e2e blocks (one per upstream)
packages/coding-agent/* defaultModelPerProvider, login, env docs,
README, providers.md
Verified end-to-end against a real Cloudflare account with unified
billing: 9/9 e2e tests pass across all three upstreams (Workers AI
Kimi K2.6, OpenAI gpt-5.1 reasoning, Anthropic claude-sonnet-4-5).
* refactor(ai): move AI Gateway User-Agent and per-route session-affinity flag to catalog
Mirrors the same per-model metadata refactor done for Workers AI in the
parent branch. All cloudflare-ai-gateway entries get the User-Agent
header. Only workers-ai/* gateway entries set
`compat.sendSessionAffinityHeaders: true` because the gateway
forwards that header to the underlying Workers AI runtime; anthropic/*
upstream and openai/* (openai-responses) don't use it.
packages/ai/scripts/generate-models.ts: emit headers (always) and
per-upstream compat (workers-ai only) on each cloudflare-ai-gateway
entry.
packages/ai/src/models.generated.ts: re-spliced 35 entries with
headers + conditional compat.
Behavior unchanged - 9/9 e2e tests pass across all three upstream
families.
* fix(ai): align AI Gateway with telemetry-aware UA helper
Adapts to badlogic/pi-mono#3851's follow-up fix ("honor telemetry for
Cloudflare attribution headers", fbb5eed) which moved the
'User-Agent: pi-coding-agent' header out of per-model catalog metadata
and into a centralized telemetry-honoring helper
(coding-agent/src/core/sdk.ts:getAttributionHeaders).
- packages/coding-agent/src/core/sdk.ts: extend the cloudflare branch of
getAttributionHeaders to also match cloudflare-ai-gateway and
gateway.ai.cloudflare.com.
- packages/ai/scripts/generate-models.ts and src/models.generated.ts:
drop 'headers' from the 35 cloudflare-ai-gateway entries (constant
CLOUDFLARE_STATIC_HEADERS no longer exists). Per-route
compat.sendSessionAffinityHeaders is unchanged.
End-to-end behavior unchanged: 9/9 tests still pass across all three
upstream families (Workers AI, Anthropic, OpenAI Responses).
---------
Co-authored-by: Mario Zechner <badlogicgames@gmail.com>
- treat tools as a global allowlist across built-in, extension, and SDK tools
- remove process-cwd singleton tool usage from SDK and CLI paths
- add regression coverage for extension tool filtering
closes#3452closes#2835
When a provider (e.g., Google Gemini CLI) requests a retry delay longer
than maxDelayMs (default: 60s), the request fails immediately with an
informative error instead of waiting silently for hours.
The error is then handled by agent-level auto-retry, which shows the
delay to the user and allows aborting with Escape.
- Add maxRetryDelayMs to StreamOptions (packages/ai)
- Add maxRetryDelayMs to AgentOptions (packages/agent)
- Add retry.maxDelayMs to settings (packages/coding-agent)
- Update _isRetryableError to match 'retry delay' errors
fixes#1123
- Add ResourceLoader interface and DefaultResourceLoader implementation
- Add PackageManager for npm/git extension sources with install/remove/update
- Add session.reload() and session.bindExtensions() APIs
- Add /reload command in interactive mode
- Add CLI flags: --skill, --theme, --prompt-template, --no-themes, --no-prompt-templates
- Add pi install/remove/update commands for extension management
- Refactor settings.json to use arrays for skills, prompts, themes
- Remove legacy SkillsSettings source flags and filters
- Update SDK examples and documentation for ResourceLoader pattern
- Add theme registration and loadThemeFromPath for dynamic themes
- Add getShellEnv to include bin dir in PATH for bash commands
Error messages for missing credentials and shell config now use
getAuthPath() and getSettingsPath() instead of hardcoded ~/.pi/agent/
paths, so they correctly reflect PI_CODING_AGENT_DIR when set.
Detect OAuth authentication failures (expired credentials, offline) and provide helpful error message instead of crashing with generic 'No API key found' error.
Co-authored-by: Mario Zechner <badlogicgames@gmail.com>
When 429/500 errors occur during tool execution, empty assistant messages
with stopReason='error' get persisted. These break the tool_use -> tool_result
chain for Claude/Gemini APIs.
Added centralized filtering in transformMessages to skip assistant messages
with empty content and no tool calls. Provider-level filters remain for
defense-in-depth.
The --no-skills flag set options.skills = [] in main.ts, but the
interactive mode UI would rediscover skills anyway because it called
loadSkills() directly instead of using the already-loaded skills.
Changes:
- Add AgentSession.skills and AgentSession.skillWarnings properties
- discoverSkills() now returns { skills, warnings } instead of Skill[]
- Interactive mode uses session.skills instead of calling loadSkills()
- Update SDK docs and examples for new return type
Fixes#577
The --no-skills flag set options.skills = [] in main.ts, but the interactive mode UI would rediscover skills anyway because it called loadSkills() directly.
Changes:
- Add AgentSession.skills and AgentSession.skillWarnings properties
- discoverSkills() now returns { skills, warnings } instead of Skill[]
- Interactive mode uses session.skills instead of calling loadSkills()
Co-authored-by: Carlos Villela <cv@lixo.org>
When passing a string systemPrompt to createAgentSession(), it is now
used as-is without appending context files and skills. This matches the
documented behavior: 'String replaces default, function receives default
and returns final.'
Previously, string systemPrompt would have context files and skills
appended, causing duplication if they were already in the string.
fixes#543
- Add setActiveTools() to ExtensionAPI for dynamic tool management
- Extensions can now override, wrap, or disable built-in tools
- Add tool-override.ts example demonstrating the pattern
- Update documentation for tool override capabilities
Tools now use ExtensionRunner.createContext() instead of a separate
inline context factory. This ensures tools and event handlers share
the same context, fixing ctx.shutdown() and other context methods.
- Made ExtensionRunner.createContext() public
- Changed wrapRegisteredTools to accept ExtensionRunner instead of getContext callback
- Create ExtensionRunner when SDK custom tools are present (not just extensions)
- Removed redundant inline context factory from sdk.ts
The tool execution context was created with a no-op shutdown handler.
Now it delegates to ExtensionRunner.shutdown() which uses the handler
set by the mode via initialize().
