You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
docs: landing page accuracy sweep — align claims with the implementation
Verified every checkable claim in docs/index.html against the code:
- JSON-LD softwareVersion 1.2.0 -> 1.26.2 (stale by 24 releases)
- Docker Compose files live in docker/, not the repo root (hero carousel
- now cds into odek/docker so the later config/compose steps resolve)
- 'go 1.24' -> 'go 1.25' in demo go.mod output (go.mod says 1.25.12)
- search_files is pure-Go regex content search, not 'ripgrep-backed'
- session_search is semantic vector search over conversation content;
there is no --deep flag
- odek serve sandboxes by default — cheatsheet showed the inverse
(--no-sandbox is the opt-out)
- odek mcp has no --sse-addr/SSE mode, stdio only — removed the cheatsheet
row and the same stale claim in llms.txt and AGENTS.md
- odek schedule add requires --cron "<expr>"
- --interaction-mode accepts a fourth mode (off); also fixed the raw <m>
tag that swallowed the flag placeholder at render time
- added the EXTENSIONS.md (odek-extension/v1) link to the docs grid and
llms.txt
Everything else checked out: install one-liner asset names, Go 1.25+,
default model/iteration count, thinking levels, learn/no-learn, repl
--id, run --deliver, memory/cleanup subcommands, telegram budget env
vars, 3-worker delegation default, network_egress=prompt default,
MIT license, Linux/macOS release targets.
Copy file name to clipboardExpand all lines: docs/index.html
+12-12Lines changed: 12 additions & 12 deletions
Original file line number
Diff line number
Diff line change
@@ -517,7 +517,7 @@
517
517
"description": "The fastest, minimal Go autonomous agent runtime. ~12 MB static binary, zero frameworks. ReAct agent with parallel tool execution, persistent memory, MCP, sandbox, Web UI, and Telegram bot.",
<p><code>session_search</code> — full-text and fuzzy search across your entire agent history. Use <code>--deep</code> for LLM-powered semantic search. Never lose a past finding, decision, or code review.</p>
967
+
<p><code>session_search</code> — semantic search across your entire agent history: vector matching over full conversation content finds sessions whose messages match your query, even when titles don't. Never lose a past finding, decision, or code review.</p>
968
968
</div>
969
969
970
970
<divclass="card">
@@ -1097,13 +1097,12 @@ <h3>Commands</h3>
1097
1097
<li><spanclass="cmd">odek run --sandbox <task></span><spanclass="desc">Run in Docker sandbox</span></li>
<ahref="https://github.com/BackendStack21/odek/blob/main/docs/CACHING.md">Prompt Caching<span>Cache markers for lower latency and cost on repeats</span></a>
1148
1147
<ahref="https://github.com/BackendStack21/odek/blob/main/docs/MCP.md">MCP (Two-Way)<span>Serve tools + connect to external MCP servers</span></a>
- [Sub-agents](https://odek.21no.de/SUBAGENTS.md): `delegate_tasks` tool and `odek subagent` — real OS processes with trust levels, risk caps, and FD-based API key handoff
19
19
- [Tool Selection](https://odek.21no.de/TOOL_SELECTION.md): restricting which built-in tools the LLM can see
20
20
- [Prompt Caching](https://odek.21no.de/CACHING.md): provider prompt-caching support, cost and latency effects
21
-
- [MCP Integration](https://odek.21no.de/MCP.md): odek as MCP server (stdio/SSE) and as MCP client, per-tool approval flow
21
+
- [MCP Integration](https://odek.21no.de/MCP.md): odek as MCP server (stdio) and as MCP client, per-tool approval flow
0 commit comments