Conversation
# Conflicts: # packages/agent/src/services/remote-plugin-adapter.test.ts # packages/cloud-shared/src/lib/services/eliza-app/onboarding-chat.ts # packages/examples/smartglasses/hardware-smoke.ts # plugins/plugin-smartglasses/src/ui/SmartglassesView.tsx
…le-merge # Conflicts: # packages/cloud-shared/src/lib/services/eliza-app/onboarding-chat.ts # packages/examples/smartglasses/hardware-evidence.ts # packages/examples/smartglasses/hardware-smoke.ts
…le-merge # Conflicts: # packages/agent/src/__tests__/plugin-tui-view-coverage.test.ts
Removes the now-deprecated REST API, Plugin Registry, Launch Resources, old CLI, examples-gallery, product/, and prototypes/ sections plus the what-you-can-build.mdx landing in favor of the new 8-track structure. Rewrites docs.json nav around the 8 tracks (Agent, Agent App, Plugin, Framework, Framework App, elizaOS, Cloud, Training & Sim) and a focused Get Started / CLI / Reference. Rewrites index.mdx to point users at the tracks rather than feature-grab-bag cards. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…taset converters fused_turboquant is incompatible with the Qwen3.5 hybrid linear+full attention architecture (18 linear + 6 full attention layers). Remove it from three places: - run_pipeline.py default quantizers + optimize_for_eliza1.py call - finetune_all_tiers.py QUANT_PIPELINE - model_registry.py quantization_after for all Qwen3.5/3.6 tiers Converter fixes: - convert_hermes_to_eliza.py: strip trope prefixes from assistant context messages so multi-turn history doesn't teach trope openings - convert_openclaw_to_eliza.py: same _strip_trope_prefix addition New: - convert_xlam_to_eliza.py: converter for Salesforce/xlam-function-calling-60k - datasets.yaml: add hermes-fc-v2, openclaw-operator, openhermes-2.5 Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds tracks/{agent,agent-app,plugin,framework,framework-app,elizaos,
cloud,training}/ with an overview and the core sub-pages per track.
Each page pulls from real packages/ source (linked) rather than
restating older marketing prose.
CLI section reduced to just `create-project` and `create-plugin`, since
that's what users actually run; everything else is project scripts.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Delete packages/docs/api-reference.mdx (REST API surface is gone)
- Drop residual packages/docs/{product,prototypes,examples,docs} dirs
- Bulk-rewrite internal links: /cli-reference -> /cli, /what-you-can-build
-> /tracks/overview, /examples-gallery -> /tracks/overview,
/plugin-registry/* -> /plugins/overview or /connectors, /rest/* dropped
- Fix discord.gg/eliza -> discord.gg/elizaos in tutorials and contributing
- Normalize ElizaOS -> elizaOS in user-facing text (keeps code identifiers)
- Update docs.test.js expected dirs to match new structure (tracks/, cli/,
connectors/ replace examples/)
All 13 docs tests pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
…rompt extraction Two bugs in the native tool-call benchmark pipeline: 1. native_tool_call_bench.py render_prompt: Qwen3.5 chat template line 120 iterates `tool_call.arguments|items`, expecting a dict. Multi-turn trajectory records have OpenAI-style messages where function.arguments is a JSON string (not yet parsed). Parse string → dict before applying the chat template. 2. benchmark_vs_cerebras.py _load_prompts: looked for record["messages"] but eliza_native_v1 format places messages at record["request"]["messages"]. Added fallback to check both locations. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
User flagged that package READMEs were drifting from code. Audited and updated the high-impact ones: - packages/core/README.md: drop phantom plugin-config helper docs (collectSettings, loadPluginConfig, etc. — never existed in src/) - packages/app/README.md: remove bogus build:desktop/dev:desktop scripts; point at the actual root-level Electrobun commands - packages/app-core/README.md: rewrite — referenced README files that don't exist (src/onboarding/README.md, connection/README.md) and a connection/ directory that doesn't exist; describe actual subdirs - packages/robot/README.md: drop stale "coming in Wave 7/8" promises; document the real scripts (robot:bridge:mock, mujoco, robot:demo) - packages/cloud-api/README.md: new — layout, dev, link to Cloud track - packages/cloud-frontend/README.md: new — dashboard surface + dev - packages/benchmarks/README.md: new — 70+ harness inventory + runner - packages/os/README.md: new — Linux + AOSP layout, variants, flashing Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
_parse_json_object used rfind("}") to find the end of the JSON object,
which broke when the model appended Qwen <tool_call> XML blocks after the
eliza JSON. Use balanced-brace walking instead so the slice ends at the
correct closing brace, ignoring any trailing XML.
Fixes 142/144 parse_errors in tool_call bucket — the model correctly
outputs eliza_native_v1 JSON tool calls; the parser was discarding them.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
gpt-oss-120b completions were truncating at 512 tokens (finish_reason=length) causing quality comparisons to score against incomplete answers. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Salesforce/xlam-function-calling-60k is now available as standard parquet — trust_remote_code is no longer supported/needed. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Delete plugins/registry.md (user said remove plugin registry) - Delete connectors/owner-vs-agent-audit.md and agents/owner-role.md (internal audit notes, not user docs) - Refine tracks/elizaos/linux.mdx + install.mdx with the real OS CLAUDE.md guidance: v1 USB-only, channels (alpha/beta/stable/Enterprise), signed-flasher requirements, release-manifest generation step, honest hardening status - Wire all 28 previously-orphan pages into the docs.json nav (lifeops-setup, developer-docs, autonomous-mode, training, wallet, emotes, themes, sandbox, local-models, media-generation, documents, triggers, registry, whatsapp, beginners-user-guide, bluebubbles, zalouser, dashboard settings/talk-mode, electrobun-startup, etc.) - Re-point dead /plugins/registry internal links at /tracks/plugin/publish Result: 188 pages, all in nav, zero orphans. 13/13 docs tests pass. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
gpt-oss-120b tool-call responses exceed 1024 tokens on complex multi-tool prompts; 2048 eliminates finish_reason=length truncations. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ents.ts
External fact-check caught drift: the EventType enum block in
runtime/events.mdx documented events that don't exist (RUN_COMPLETED,
RUN_FAILED, MESSAGE_UPDATED, ROOM_UPDATED, VOICE_STARTED, VOICE_ENDED,
ACTION_FAILED, EVALUATOR_FAILED, MODEL_FAILED, SERVICE_STARTED/STOPPED/
ERROR) and used colon-case "message:received" strings when the real
enum values are uppercase "MESSAGE_RECEIVED".
Rewrite the enum block verbatim from packages/core/src/types/events.ts:
- Drop the phantom events.
- Add the missing ones: CHANNEL_CLEARED, REACTION_RECEIVED, POST_GENERATED,
INTERACTION_RECEIVED, EMBEDDING_GENERATION_{REQUESTED,COMPLETED,FAILED},
CONTROL_MESSAGE, FORM_FIELD_{CONFIRMED,CANCELLED}, HOOK_* (15 hooks),
PIPELINE_HOOK_METRIC, DISCORD/TELEGRAM/X platform channels.
- Fix the handler-registration example: RUN_COMPLETED -> RUN_ENDED.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The /bsc promo and any other direct-crypto-payments surface require `user.wallet_address` to be set. That field is only assigned when a user signs in via SIWE/SIWS — Google / Discord / GitHub / Magic Link / Passkey signups never get one, and there was no path to add one later. The pay button on /bsc was permanently dead for those accounts. Add `POST /api/users/me/wallet/attach`: takes a SIWE message + signature, validates against the existing nonce store, asserts the address isn't already bound to a different user, then writes wallet_address / wallet_chain_type / wallet_verified onto the authed user. The repository update method already exists; we just route it. Conflict policy: a wallet bound to another account returns 409 `wallet_taken`. The wallet-keyed account wins; OAuth users hitting this must use the existing account. UI: when the authed user has no wallet_address, /bsc swaps the purchase card for an AttachWalletCard that drives connect -> sign -> POST attach -> invalidate user-profile. After the mutation succeeds, the page re-renders with the existing DirectCryptoCreditCard unchanged. Coverage: - 6 backend unit tests on the route (94% line coverage), covering already_attached / 400 / 401 / wallet_taken / happy path / same-user race window. - 3 Playwright e2e tests on /bsc covering purchase, OAuth-no-wallet attach, and HTML-fallback safety. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Contributor
|
Important Review skippedAuto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Contributor
|
Claude encountered an error —— View job I'll analyze this and get back to you. |
Contributor
LifeOps Multi-Tier BenchmarkSuite:
|
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Promote
develop(36 commits) tomain. The user-visible win:feat(cloud): let OAuth users attach a wallet to buy crypto credits(74d23ff6e4) — fixes the /bsc dead-end where OAuth-signed-in users (Google / Discord / GitHub / Magic Link / Passkey) couldn't pay becauseuser.wallet_addresswas unset and there was no path to set it. AddsPOST /api/users/me/wallet/attach(SIWE-verified) and an "Verify your BSC wallet" card on /bsc that drives connect → sign → attach → invalidate before handing off to the existing purchase flow. 6 unit tests on the route + 3 Playwright e2e tests on /bsc.Other commits in this batch: doc tidies, benchmark fixes (Cerebras max_tokens, JSON extraction for Qwen, xlam dataset trust flag), runtime EventType doc alignment.
Test plan
74d23ff6e4).https://elizacloud.ai/bscwhile signed in via Google: the page now shows "Verify your BSC wallet" instead of the dead Pay button.🤖 Generated with Claude Code