Update README: public repo, model-agnostic inference (remove DeepSeek/local-only framing)
This commit is contained in:
@@ -5,8 +5,8 @@
|
|||||||
|
|
||||||
## Repo
|
## Repo
|
||||||
|
|
||||||
- **Location:** `Leonard/athena-oracle` (private, Gitea)
|
- **Location:** `Tony_tech/athena-oracle` (public, Gitea)
|
||||||
- **URL:** http://localhost:3000/Leonard/athena-oracle
|
- **URL:** http://localhost:3000/Tony_tech/athena-oracle
|
||||||
- **Branch:** `main`
|
- **Branch:** `main`
|
||||||
- **Origin:** mirrors `~/oracle` (local working copy)
|
- **Origin:** mirrors `~/oracle` (local working copy)
|
||||||
|
|
||||||
@@ -23,15 +23,25 @@ momentum rather than source-expansion noise.
|
|||||||
| Theme scan | `theme_scan.py` | Cross-source trend detection + idempotent falsification |
|
| Theme scan | `theme_scan.py` | Cross-source trend detection + idempotent falsification |
|
||||||
| Query | `query.py` | Interactive lookup against the store |
|
| Query | `query.py` | Interactive lookup against the store |
|
||||||
| Archive | `archive.py` | Cold-storage rotation |
|
| Archive | `archive.py` | Cold-storage rotation |
|
||||||
| Summarize | `summarize.py` | Local summarization (Ollama — currently offline) |
|
| Summarize | `summarize.py` | Summarization via any available inference model |
|
||||||
| Schema | `schema.sql` | SQLite store definition |
|
| Schema | `schema.sql` | SQLite store definition |
|
||||||
| Cron entry | `oracle-pipeline.sh` | Wrapper invoked by Hermes cron |
|
| Cron entry | `oracle-pipeline.sh` | Wrapper invoked by Hermes cron |
|
||||||
|
|
||||||
## Local inference status
|
## Inference model strategy
|
||||||
|
|
||||||
`summarize.py` uses Ollama only (DeepSeek fallback removed — OpenRouter 402).
|
Athena is **model-agnostic** — it uses whatever inference backend is available at
|
||||||
Ollama is **currently unavailable**; summarization steps are skipped until it returns.
|
run time, whether free or paid. There is no hard dependency on a single provider.
|
||||||
Pipeline ingestion, scoring, and theme-scan continue without it.
|
|
||||||
|
`summarize.py` currently targets a local Ollama endpoint (`llama3.2:1b`) when
|
||||||
|
present. The pipeline is designed so the summarization backend can be swapped for
|
||||||
|
any model we can reach — local GPU, a paid API, or a free-tier endpoint — without
|
||||||
|
changing the ingestion, scoring, or theme-scan logic. When no inference backend is
|
||||||
|
reachable, the summarization step is skipped; ingestion, scoring, and theme-scan
|
||||||
|
continue uninterrupted.
|
||||||
|
|
||||||
|
To wire in a different backend, implement the same `summarize(text) -> (summary, model)`
|
||||||
|
contract that `summarize_with_ollama` satisfies, and add the dispatch in
|
||||||
|
`process_card`.
|
||||||
|
|
||||||
## Data handling
|
## Data handling
|
||||||
|
|
||||||
|
|||||||
+1
-1
@@ -2,7 +2,7 @@
|
|||||||
|
|
||||||
## White Paper — System Architecture, Pattern Detection, and Opportunity Discovery
|
## White Paper — System Architecture, Pattern Detection, and Opportunity Discovery
|
||||||
|
|
||||||
**Source:** `Leonard/athena-oracle` (private, Gitea — http://localhost:3000/Leonard/athena-oracle) · branch `main`
|
**Source:** `Tony_tech/athena-oracle` (public, Gitea — http://localhost:3000/Tony_tech/athena-oracle) · branch `main`
|
||||||
|
|
||||||
---
|
---
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user