Index, TandemLLM v0.13.0 documentation

A B C D E F G H I J K L M N O P Q R S T U V W X Z

Each entry names the chapters where the term appears; the link goes to the place in the text.

A

adapter: Glossary
admin password: 3
admission: 14
AES-256-GCM: 5
Agents mode: 28
AGPL-3.0-only: About this manual
anchor: 2, 13, Glossary
--answer-reserve: 14
auth: 32

B

backend: Glossary
backend kind: 5
--backends: 2, 4, 26, 27, 32
Backends view: 27
benchmark row: 15, Glossary
block drafter: 11, Glossary
board: Glossary
byte budget: 8

C

cache exactness: 12
cache source: Glossary
cache tiers: 13
CC BY 4.0: About this manual
chatd: 28
Chat tab: 28
chat template: 14
claim: Glossary
claim check: 23
clip search: 9
commit: 11, Glossary
community weights: 7
community weights, Super VL: 7
community weights, sync: 7
compaction: 13, 13, Glossary
conductor: 21, Glossary
--conductor-config: 3, 26
conductor documents: 21, 31
Conductor view: 27
context length: 11
cost: 5
cron: 6
CUDA graphs: 10

D

dashboard: 4
dashboard contract: 27
dashboard login: 3
decider: 22, Glossary
decode loop: 8
DeepSeek: 5
deploy: 6
DFlash2: 11
DGX Spark: 1
--disk-cache: 3, 13
disk tier: 13
docker: 4
drafter: 1, 11, Glossary
draft tree: 11, Glossary
drain: 6
--drop-idle: 11

E

effort: 21, Glossary
end anchor: 2, 13, Glossary
--end-anchor: 13
engine/cache.py: 14
environment overlay: 5
errors: 32
EXL3: 9
EXL3 pack: 9, 17
EXL3, standalone pack: 9
EXL3, the 27B: 9

F

fallback: 5
feature group: 26, Glossary
Flash-Next: 17
Flash-Next, exactness: 12
Flash-Next, measurement: 15
Flash-Next, memory notes: 25
Flash-Next, operations: 6
Flash-Next, serving: 14
Flash-Next, state: 13
FP8: 1, 9

G

Gated DeltaNet: 8, 10
GB10: 1
GPTQ: 9
Grafana: 6
greedy agreement: 1, 12
grid: 13, Glossary
guest: 13, Glossary
Gumbel-max trick: 12

H

harness: Glossary
headers: 32
Hermes agent: 2
hold: 6, Glossary
host collector: 29, Glossary
Hugging Face: 3, 9

I

images: 14
install.sh: 3

J

journal: 21, Glossary

K

keychain: 5
keys at rest: 5
KV rows: 8

L

L1, L2: Glossary
Laya: 1, 22
Laya API: 22
length router: 11
LFM2.5: 18
library: 22
license: About this manual
LiLiCorr: 19
Live panel: 27
llama.cpp: 5
LM Studio: 5
logs: 6
lookup drafter: 11, Glossary
loop guard: 23

M

master key: 5
--max-image-rows: 14
--max-images: 14
measured at the proxy: 5
--mem-ceiling-gib: 14
memory gate: 14, Glossary
memory note: 25, Glossary
memory rules: 6
memory writer: 25, Glossary
--mem-peak-gib: 14
metrics: 33
model adapter: 16
ModelState: 16
monitoring: 6
MTP drafter: 11

N

narrator: 1, 24, Glossary
near tie: 12, Glossary
Nemotron 3.5 Bolt: 19
Nemotron 3.5 Super VL: 19
never worse: 1, 15
--no-learn-cost: 11, 15, 24, 37
note: 21, Glossary
NVFP4: 1, 9

O

Ollama: 5
one engine per board: 3, 6
OpenAI: 5
OpenAI API: 14
OpenAI SDK: 2
opencode: 2
Open WebUI: 2
ops/gate.sh: 15, 16
ops/hold.sh: 6
ops/serve.env: 3, 6, 10, 11, 30, 31
ops/start.sh: 3, 6, 21, 30, 31
ops/supervisor.yaml: 23
Ornith-1.5-122B: 20
outcome: Glossary

P

package: 2, 22, Glossary
Packages screen: 22
page cache: 6
--parallel: 14
parallel requests: 14
Parallel view: 29
park: 13, Glossary
permission modes: 28
pin: 13, Glossary
plugin: 29, Glossary
policy: 23
prefill: 13
--prefix-chunk: 13
privacy: 14
profile: 2, 21, Glossary
profile (serve.env): 3
Prometheus: 6, 33
proxy: 1, 4, Glossary

Q

QSE_ADMIN_TOKEN: 32
QSE_BACKENDS: 3, 4, 31
QSE_CONDUCTOR_CONFIG: 3, 31
QSE_DASHBOARD_LOGIN: 31
QSE_DISK_CACHE: 3, 31
QSE_USAGE_LEDGER: 3, 31
QSE variables: 31
quality gate: 9
quantiser: 9
--queue-timeout: 14
Qwen3.8-27B: 8
QWEN38_DECIDER_MODELS: 22
QWEN38_DEEP: 11
QWEN38_NARRATOR_MODEL: 24
QWEN38_SHADOW_MODEL: 24
QWEN38_STAIR_RHO: 11
QWEN38_STAIR_TABLES: 11
QWEN38 variables: 31

R

raw: Glossary
reading line: 21
reasoning budget: 11
reasoning format: 14
--reasoning-format: 14
recurrent state: 8
release gate: 15
release identity: 12
replay: 21
--request-timeout: 14
requirements: 3
--reserve-slack-gib: 14
residency planner: 3, 6, Glossary
--resident-gb: 13
--resident-guest-max: 13
--resident-park-gb: 13
--resident-pin-gb: 13
resident prefix: 2, 13, Glossary
--response-cache: 12
Responses API: 5
role: 6, 22, Glossary
rollback: 6
round: 11, Glossary
router: 21, Glossary
routing by model id: 5
row invariance: 10
run-to-run text: 12

S

safe point: 23, Glossary
sampling: 11
sandbox: 28
seed: 12
self-check: 17
serve.env: 31
server/app.py: 14
server flags: 30
session address: 21
session monitor: 2, 23, Glossary
session resume: 12, Glossary
session spill: 13, Glossary
--session-spill: 13
settings registry: 6
setup mode: 3
setups: 3
Setup screen: 4
SGLang: 5
shadow mode: Glossary
shadow reader: 1, 24, Glossary
--side-graphs: 14
side lane: 14, Glossary
side request: 21, Glossary
sign-in: 3
skinny kernel: 10
spill: Glossary
SSE comment: 14
staircase: 11
StairCut: 1, 11, 11, Glossary
StairCut, Flash-Next: 11
StairCut paper: 1
start.sh: 6
state store: 13, Glossary
status line: 24, Glossary
status page: 6
stop.sh: 6
store view: 15
streaming: 14
structured outputs: 14
suffix store: 11, 13, Glossary
systemd: 6
System view: 27

T

taint: Glossary
TANDEM_API_KEYS: 5
TANDEM_COLLECT_TOKEN: 29
tandem-dash: 4, 30
tandem-dash hostcollect: 4
tandem-dash launch: 5, 5
TandemLLM: 1
TandemLLM (model name): 2, 21
TANDEM_SECRETS_KEY: 5
TANDEM variables: 31
task class: 21, Glossary
teacher-forced bench: 15
test environment: 1
thinking budget: 14
time to first token: 13, Glossary
tokens per round: 8
tool-argument loop guard: 14
tool-call checks: 23
tool calls: 14
tools/block_ab.py: 15
tools/forced_bench.py: 15
tools/row3.py: 15
tools/verify_spec.py: 12, 15, 16
tree verify: 11
--trust-loopback: 32
turn outcome: 23

U

ulp: 12, Glossary
uninstall: 3
upgrade: 3, 3
usage ledger: 33
--usage-ledger: 3, 33
Usage view: 27
use-profile.sh: 3
users: 5

V

verify: 11, Glossary
vision tower: 14
vLLM: 5
voice: 24, Glossary

W

W4A16: 9
waiting line: 14
watchdog: 6
weight profile: 6
workers: 28

X

X-OpenWebUI-Chat-Id: 21
x-session-id: 21
X-Tandem: 3, 21, 23
X-Tandem header: 3
X-Tandem-Profile: 3, 21

Z

Zenodo: 1