Tony_tech/athena-oracle
3
0
Fork 0
Code Issues 6 Pull Requests 2 Actions Packages Projects Releases Wiki Activity

Inference backend idle — Ollama not running, no model #8

New Issue
Closed
opened 2026-07-10 15:39:28 +00:00 by Leonard · 2 comments
Leonard commented 2026-07-10 15:39:28 +00:00
Copy Link
Copy Source

Category: Infra · Priority: LOW

Summary
ollama binary is installed (~/.local/bin/ollama) but the process is not running and no model is pulled. Blocks the planned LLM features (MoA claim-check #5, semantic verify, embeddings #6).

Why it matters
Both roadmap upgrades (embeddings, MoA) need a running inference backend on this CPU-only, 2-vCPU, 1.8GB-RAM box.

What happens if not done
Roadmap upgrades can't ship; Athena stays LLM-free. Acceptable today (pipeline is deterministic by design) but blocks later work.

Suggested approach
Decide trigger: when #6 lands, start Ollama per cron-run (not 24/7, given RAM headroom) and pull nomic-embed-text (~274MB). Check ~/.ollama (already 1.3GB) — a model may already be present.

**Category:** Infra · **Priority:** LOW **Summary** `ollama` binary is installed (`~/.local/bin/ollama`) but the process is not running and no model is pulled. Blocks the planned LLM features (MoA claim-check #5, semantic verify, embeddings #6). **Why it matters** Both roadmap upgrades (embeddings, MoA) need a running inference backend on this CPU-only, 2-vCPU, 1.8GB-RAM box. **What happens if not done** Roadmap upgrades can't ship; Athena stays LLM-free. Acceptable today (pipeline is deterministic by design) but blocks later work. **Suggested approach** Decide trigger: when #6 lands, start Ollama per cron-run (not 24/7, given RAM headroom) and pull `nomic-embed-text` (~274MB). Check `~/.ollama` (already 1.3GB) — a model may already be present.
Leonard commented 2026-07-12 16:33:54 +00:00
Author
Copy Link
Copy Source

Closed — decision documented. Ollama is NOT needed: summarization moved to sub-agent cron (athena-summarize-batch, 4h) using the Hermes model (Qwen 3.6 @ 100.64.0.2:39195). This is better than Ollama on this hardware (CPU-only, 2 vCPU, 1.8GB RAM) — no local model overhead, higher quality summaries, triage by signal_score so only the top 20 per batch get summarized. Ollama binary remains installed but unused.

Closed — decision documented. Ollama is NOT needed: summarization moved to sub-agent cron (`athena-summarize-batch`, 4h) using the Hermes model (Qwen 3.6 @ 100.64.0.2:39195). This is better than Ollama on this hardware (CPU-only, 2 vCPU, 1.8GB RAM) — no local model overhead, higher quality summaries, triage by signal_score so only the top 20 per batch get summarized. Ollama binary remains installed but unused.
Leonard commented 2026-07-12 16:41:16 +00:00
Author
Copy Link
Copy Source

Closed — decision documented. Ollama is NOT needed: summarization moved to sub-agent cron (athena-summarize-batch, 4h) using the Hermes model (Qwen 3.6). This is better than Ollama on this hardware (CPU-only, 2 vCPU, 1.8GB RAM). Ollama binary remains installed but unused.

Closed — decision documented. Ollama is NOT needed: summarization moved to sub-agent cron (athena-summarize-batch, 4h) using the Hermes model (Qwen 3.6). This is better than Ollama on this hardware (CPU-only, 2 vCPU, 1.8GB RAM). Ollama binary remains installed but unused.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
No items
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
HA_Researcher Ty tonyjbala (Tony Balascio)
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Tony_tech/athena-oracle#8
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?