S
{{ item.glyph }}
◐
{{ screenTitle }}
sa8775-expert ▾
+ New session
Recent
Max junction temperature…
open · 2 min ago
PMIC power sequencing
yesterday
Thermal derating curves
2 days ago
GPIO drive strength options
4 days ago
What is the max junction temperature for the SA8775?
S
The maximum junction temperature is 105 °C for the commercial grade SA8775 TRM · p.87 §2.4.1 and 125 °C for the automotive grade (AEC-Q100 Grade 2) SA8775 TRM · p.88 §2.4.2.
● grounded 0.94 — view evidence qwen2.5:32b · 2.3 s copy
What's the max junction temp of the SA8797?
⊘
Not answering — retrieval too weak best match 0.22 · floor 0.35
The knowledge base doesn't appear to cover thermal limits for the SA8797 variant, so this instance won't guess. Nearest material found:
0.22SA8775 TRM · p.87 · thermal, commercial grade
0.19SA8775 TRM · p.88 · thermal, automotive grade
0.14SA8775 TRM · p.91 · power dissipation limits
Ask anyway (ungated)
Adjust gate
Add sources to KB
How should PMIC power sequencing be ordered at cold boot?
S
thinking·retrieving (8)·generating·verifying
Cold-boot sequencing brings rails up in dependency order: VDD_CORE first, held stable for 2 ms before
{{ composerHint }}
+
gate ≥0.85 · refuse
sa8775-docs@latest
↑
Grounded in your sources · refuses when unsupported · nothing leaves this machine
Evidence
grounded 0.94
✕
Claims · 2 of 2 supported
Commercial grade TJ max is 105 °C
→ chunk #4411 · 0.91
Automotive grade (AEC-Q100 G2) TJ max is 125 °C
→ chunk #4416 · 0.88
Retrieved chunks
SA8775 TRM›2.4.1 Commercial0.91
…under continuous operation the maximum junction temperature T J shall not exceed 105 °C for commercial grade devices, derated per Figure 2-9…
p.87 · sa8775-docs@a3f9c21
SA8775 TRM›2.4.2 Automotive0.88
…qualified to AEC-Q100 Grade 2; automotive grade parts permit a maximum junction temperature of 125 °C with mission-profile derating…
p.88 · sa8775-docs@a3f9c21
SA8775 TRM›2.5 Power limits0.61
…total platform dissipation budget of 23 W assumes heatsink attach per §9.2 and ambient below 85 °C…
p.91 · sa8775-docs@a3f9c21
Instances
Specialized, private AI instances grounded in your knowledge bases.
New instance
SA
sa8775-expert
ollama · qwen2.5:32b-instruct · memory on · used 2 min ago
sa8775-docs@latest ≥0.85 refuse Chat →
AU
aou-auditor
ollama · qwen2.5-coder:32b · memory off · used 3 h ago
firmware-src aou-docs ≥0.90 refuse Chat →
TW
test-writer
ollama · qwen2.5-coder:32b · memory on · used 1 week ago
team-standards@latest ≥0.70 hedge Chat →
← All instances
sa8775-expert
form
yaml
Save
# instances/sa8775-expert.yaml — edits here sync with the form
name: sa8775-expert
provider: ollama
model: qwen2.5:32b-instruct
system_prompt: |
  You are an expert on the SA8775 automotive platform.
  Answer only from the mounted knowledge base.
knowledge_bases:
  - name: sa8775-docs
    snapshot: latest
accuracy_gate:
  min_grounding: 0.85
  below_threshold: refuse
  require_citations: true
  retrieval_floor: 0.35
  self_consistency: false  # slow
memory:
  enabled: true
  summarize_every: 8
  token_budget: 2000
generation:
  temperature: 0.2
  max_tokens: 2048
Identity
Name
sa8775-expert
Provider
ollama▾
Model
qwen2.5:32b-instruct▾
Accuracy gate ≥0.85 · refuse
Min grounding score0.85
Share of claims that must be supported by retrieved sources.
Action below threshold
refuse
hedge
warn
Require citations
Retrieval floor0.35
Self-consistencyslow
System prompt
You are an expert on the SA8775 automotive platform.
Answer only from the mounted knowledge base.
Knowledge bases
sa8775-docs pin: latest ▾ unmount
+ Mount a knowledge base
Memory
Enabled
Summarize every8 turns
Token budget2000
Generation
Temperature
0.2
Max tokens
2048
Knowledge Bases
An ingested, versioned corpus. Instances mount it; answers cite it.
New knowledge base
sa8775-docs
4 docs · 3,011 pages
18,442 chunks · 3 snapshots
mounted by 1 instance
firmware-src
1,204 files
96,110 chunks · 6 snapshots
mounted by 1 instance
aou-docs
12 docs · 340 pages
2,881 chunks · 2 snapshots
mounted by 1 instance
team-standards
7 docs · 88 pages
610 chunks · 1 snapshot
mounted by 1 instance
sa8775-docs @a3f9c21
sources & ingest
snapshots
sa8775-trm.pdf3,011 pagesingesting
sa8775-pmic-app-note.pdf214 pages · 1,916 chunksindexed
sa8775-thermal-design.pdf96 pages · 804 chunksindexed
aec-q100-summary.pdf41 pages · 388 chunksindexed
ingesting sa8775-trm.pdfstage 2 / 3
parse
✓
chunk
40%
embed
queued
› chunked sa8775-trm.pdf p.1200/3011 · 7,412 chunks so far
Drop files or directories here — or browse
a3f9c21 ⧉ Jul 2, 14:11 chunk 512/64 nomic-embed-text +2 files · +2,304 chunks latest · pinned by 1 instance
8be0d44 ⧉ Jun 18, 09:42 chunk 512/64 +1 file · +1,916 chunks roll back delete
2c11f7a ⧉ Jun 4, 17:03 chunk 384/48 initial ingest · 2 files roll back delete
Snapshots are immutable. Instances pin latest or a specific hash; rolling back never rewrites history.
Memory
What this instance carries between sessions — inspect it, edit it, wipe it.
sa8775-expert▾
Pinned factsuser-stated · always in context+ Pin a fact
pinned Our target is the SA8797 rev B variant edit unpin delete
Session summariesauto-generated
Discussed PMIC power sequencing; user cares about cold-boot timing.
from session #12 · yesterday
Sessions
#14Max junction temperature…open2m ago
#12PMIC power sequencingclosed · yesterday
#11Thermal derating curvesclosed · 2d ago
Wipe memory
Deletes all pinned facts, summaries, and transcripts for this instance. Unrecoverable.
Wipe…
Wipe memory for sa8775-expert?
This deletes 1 pinned fact, 1 summary, and 14 session transcripts. Rows are deleted and the database vacuumed. There is no undo and no backup.
Type the instance name to confirm
sa8775-expert
Cancel
Wipe memory
Models
What's available on your providers. Pulls stream layer-by-layer.
Pull model
ollama healthy
openai-compat not configured
modelfamilyparameterssizeused by
qwen2.5:32b-instructqwen232.8B19 GBsa8775-expert
qwen2.5-coder:32bqwen232.8B19 GBaou-auditor +1
llama3.1:8bllama8.0B4.7 GB—
nomic-embed-textnomic-bert137M274 MBembeddings default
Pull model
✕
qwen2.5:72b-instruct
pulling…
manifest
✓
6e4c87b3
✓ 12.1 GB
a91f7d02
17.3 / 28.0 GB
c22b0e91
queued
› 41 MB/s · about 4 min remaining
Downloads go to the local provider only. Nothing is sent anywhere.
Settings
Global server config. Everything here is ~/.lore/config.yaml underneath.
form
yaml
# ~/.lore/config.yaml
server:
  host: 127.0.0.1
  port: 5673
providers:
  ollama:
    url: http://127.0.0.1:11434
  openai_compat:
    base_url: null
    api_key_env: LORE_OPENAI_KEY  # read from env, never stored
embeddings:
  default_model: nomic-embed-text
log_level: info
Server
Host
127.0.0.1
Port
5673
Providers
Ollama URL
http://127.0.0.1:11434
OpenAI-compatible base URL
https:// … (optional)
API key is read from LORE_OPENAI_KEY at runtime — never stored in config or on disk.
Embeddings & logging
Default embedding model
nomic-embed-text▾
Log level
info▾
DoctorRun again
✓ollama reachable at 127.0.0.1:1143412 ms
✓4 models present, all instance models resolvable
✓GPU detectedRTX 4090 · 24 GB VRAM
✓disk space for KB index312 GB free
✗openai-compat endpoint not configuredoptional