Cloudflare Workers AI / AI Gateway resolved provider config from the
credential only, never consulting ambient env for a field the credential
omitted. The coding-agent /login flow stores just the API key, so
CLOUDFLARE_ACCOUNT_ID lives only in the environment; the key-only credential
short-circuited the env lookup, the account id stayed unresolved, and requests
hit the literal {CLOUDFLARE_ACCOUNT_ID} base URL -> 404.
resolveValue now merges per field: prefer the credential value, fall back to
ctx.env(name).
closes#6021
Signed-off-by: Mark Phelps <209477+markphelps@users.noreply.github.com>
Co-authored-by: Mario Zechner <badlogicgames@gmail.com>
Do not treat Pi’s internal ambient-auth marker as a Bedrock bearer token. This preserves SigV4 signing for AWS profiles, IAM credentials, and roles while retaining bearer authentication for real Bedrock API keys.\n\nFixes #6531
This adds cache-friendly dynamic tool loading anchored to tool results. Purely additive active-tool changes are recorded with `addedToolNames`, allowing supported Anthropic and OpenAI Responses models to load tool definitions at the point they become available instead of placing them in the cached prompt prefix.
It retains safe fallback behavior for unsupported models and non-additive changes but it will wipe caches.
getShellPath() now runs the configured shellPath through the existing
normalizePath() helper, mirroring getSessionDir(). This lets shellPath
point at a home-directory-relative shell wrapper (e.g.
~/.local/bin/agent-shell-sandbox) and resolves consistently across
machines and OSes with different home directories.
refs #6458
Detect prompt cache misses per turn by comparing each assistant message's
cache reads against the previous request's prompt tokens (core/cache-stats.ts).
Significant misses emit a warning-colored transcript notice at the turn they
occur, noting idle gaps past the cache TTL and model switches when relevant.
/session gained cache statistics: a compact token/cache breakdown with hit
rate, a $-prefixed cost section with per-model cost breakdown, and the
cumulative cost re-billed due to cache misses.
Updates GitHub Copilot built-in model metadata so models with GitHub's extended 1M capability use contextWindow 1000000 in Pi, preventing early compaction and under-budgeting for Claude Opus 4.7/4.8 and GPT-5.3 Codex/5.4/5.5.
Verified against raw docs markdown extended-capabilities table and live /models reports.
closes#6439
Export both session storage implementations from @earendil-works/pi-agent-core
so they can be imported and extended directly instead of via repo workaround.
closes#6435
in cases where forking a session was not immediate due to extensions
slowing down teardown, it's possible for the user to select the
message to fork more than once and then create multiple session forks.
in this fix we close the forking menu before starting the fork process
to not allow such behavior.
fixes#6321
* copy native clipboard .node files to the clipboard package
under the bun binary packaging, require("@mariozechner/clipboard-linux-x64-gnu")
doesn't work so instead we rely on require("./clipboard.linux-x64-gnu.node")
for that we need to copy the .node files to the clipboard package dir
* fallback to xclip if native clipboard fails on x11
Allow callers to attach an opaque JSON object to the session header so
application context needed to rebuild a harness (for example an agent
profile reference) is readable from the first line alone, without
scanning session entries. The field is optional and ignored by readers
that do not use it; fork inherits the source metadata unless overridden.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Use the new provider-specific Xiaomi Token Plan catalogs from models.dev instead of cloning the API-billing Xiaomi catalog into every token-plan region. This removes API-billing-only models such as mimo-v2-omni from xiaomi-token-plan-{cn,ams,sgp}.
* fix(ai,agent): stop salvaging malformed tool-call argument JSON
Finalized tool-call arguments must strict-parse (allowing only lossless
string-escape repair). When the streamed JSON is truncated or malformed,
the raw JSON is preserved on ToolCall.malformedArguments, arguments
become {}, and the agent loop refuses to execute the call with an error
tool result instead of running it with silently salvaged partial
arguments. Partial parsing is now only used for streaming previews.
* fix(agent): fail tool calls from length-truncated assistant messages
Reverts the malformed-arguments tracking approach in favor of handling
truncation in the agent loop: when an assistant message stops with
"length", all tool calls in it are potentially borked (streamed
arguments are salvage-parsed), so none are executed. Each gets an error
tool result telling the model to re-issue the call, and the loop
continues. No new fields on ToolCall and no provider changes.
fixes#6284
Alexey Zaytsev